├── LICENSE.md
├── Prior_Predictive_Sampling_Toggle_Switch
├── Makefile
├── run_bench.sh
├── toggle_switch_ABC_pps_naive.R
├── toggle_switch_ABC_pps_opt.R
└── toggle_switch_ABC_pps_vec_par.c
├── Prior_Predictive_Sampling_Tuberculosis
├── Makefile
├── run_bench.sh
└── tuberculosis_ABC_pps_vec_par.c
├── README.md
├── SMC_Inference_BEGE_model
├── Makefile
├── MonthlyReturns2018.csv
├── SMC_RW_LikeAnneal_BEGE_novec_nopar.c
├── SMC_RW_LikeAnneal_BEGE_novec_par.c
├── SMC_RW_LikeAnneal_BEGE_vec_nopar.c
├── SMC_RW_LikeAnneal_BEGE_vec_par.c
└── run_bench.sh
└── Weakly_Informative_Priors
├── Makefile
├── run_bench.sh
└── weak_info_test_vec_par.c
/LICENSE.md:
--------------------------------------------------------------------------------
1 | ### GNU GENERAL PUBLIC LICENSE
2 |
3 | Version 3, 29 June 2007
4 |
5 | Copyright (C) 2007 Free Software Foundation, Inc.
6 |
7 |
8 | Everyone is permitted to copy and distribute verbatim copies of this
9 | license document, but changing it is not allowed.
10 |
11 | ### Preamble
12 |
13 | The GNU General Public License is a free, copyleft license for
14 | software and other kinds of works.
15 |
16 | The licenses for most software and other practical works are designed
17 | to take away your freedom to share and change the works. By contrast,
18 | the GNU General Public License is intended to guarantee your freedom
19 | to share and change all versions of a program--to make sure it remains
20 | free software for all its users. We, the Free Software Foundation, use
21 | the GNU General Public License for most of our software; it applies
22 | also to any other work released this way by its authors. You can apply
23 | it to your programs, too.
24 |
25 | When we speak of free software, we are referring to freedom, not
26 | price. Our General Public Licenses are designed to make sure that you
27 | have the freedom to distribute copies of free software (and charge for
28 | them if you wish), that you receive source code or can get it if you
29 | want it, that you can change the software or use pieces of it in new
30 | free programs, and that you know you can do these things.
31 |
32 | To protect your rights, we need to prevent others from denying you
33 | these rights or asking you to surrender the rights. Therefore, you
34 | have certain responsibilities if you distribute copies of the
35 | software, or if you modify it: responsibilities to respect the freedom
36 | of others.
37 |
38 | For example, if you distribute copies of such a program, whether
39 | gratis or for a fee, you must pass on to the recipients the same
40 | freedoms that you received. You must make sure that they, too, receive
41 | or can get the source code. And you must show them these terms so they
42 | know their rights.
43 |
44 | Developers that use the GNU GPL protect your rights with two steps:
45 | (1) assert copyright on the software, and (2) offer you this License
46 | giving you legal permission to copy, distribute and/or modify it.
47 |
48 | For the developers' and authors' protection, the GPL clearly explains
49 | that there is no warranty for this free software. For both users' and
50 | authors' sake, the GPL requires that modified versions be marked as
51 | changed, so that their problems will not be attributed erroneously to
52 | authors of previous versions.
53 |
54 | Some devices are designed to deny users access to install or run
55 | modified versions of the software inside them, although the
56 | manufacturer can do so. This is fundamentally incompatible with the
57 | aim of protecting users' freedom to change the software. The
58 | systematic pattern of such abuse occurs in the area of products for
59 | individuals to use, which is precisely where it is most unacceptable.
60 | Therefore, we have designed this version of the GPL to prohibit the
61 | practice for those products. If such problems arise substantially in
62 | other domains, we stand ready to extend this provision to those
63 | domains in future versions of the GPL, as needed to protect the
64 | freedom of users.
65 |
66 | Finally, every program is threatened constantly by software patents.
67 | States should not allow patents to restrict development and use of
68 | software on general-purpose computers, but in those that do, we wish
69 | to avoid the special danger that patents applied to a free program
70 | could make it effectively proprietary. To prevent this, the GPL
71 | assures that patents cannot be used to render the program non-free.
72 |
73 | The precise terms and conditions for copying, distribution and
74 | modification follow.
75 |
76 | ### TERMS AND CONDITIONS
77 |
78 | #### 0. Definitions.
79 |
80 | "This License" refers to version 3 of the GNU General Public License.
81 |
82 | "Copyright" also means copyright-like laws that apply to other kinds
83 | of works, such as semiconductor masks.
84 |
85 | "The Program" refers to any copyrightable work licensed under this
86 | License. Each licensee is addressed as "you". "Licensees" and
87 | "recipients" may be individuals or organizations.
88 |
89 | To "modify" a work means to copy from or adapt all or part of the work
90 | in a fashion requiring copyright permission, other than the making of
91 | an exact copy. The resulting work is called a "modified version" of
92 | the earlier work or a work "based on" the earlier work.
93 |
94 | A "covered work" means either the unmodified Program or a work based
95 | on the Program.
96 |
97 | To "propagate" a work means to do anything with it that, without
98 | permission, would make you directly or secondarily liable for
99 | infringement under applicable copyright law, except executing it on a
100 | computer or modifying a private copy. Propagation includes copying,
101 | distribution (with or without modification), making available to the
102 | public, and in some countries other activities as well.
103 |
104 | To "convey" a work means any kind of propagation that enables other
105 | parties to make or receive copies. Mere interaction with a user
106 | through a computer network, with no transfer of a copy, is not
107 | conveying.
108 |
109 | An interactive user interface displays "Appropriate Legal Notices" to
110 | the extent that it includes a convenient and prominently visible
111 | feature that (1) displays an appropriate copyright notice, and (2)
112 | tells the user that there is no warranty for the work (except to the
113 | extent that warranties are provided), that licensees may convey the
114 | work under this License, and how to view a copy of this License. If
115 | the interface presents a list of user commands or options, such as a
116 | menu, a prominent item in the list meets this criterion.
117 |
118 | #### 1. Source Code.
119 |
120 | The "source code" for a work means the preferred form of the work for
121 | making modifications to it. "Object code" means any non-source form of
122 | a work.
123 |
124 | A "Standard Interface" means an interface that either is an official
125 | standard defined by a recognized standards body, or, in the case of
126 | interfaces specified for a particular programming language, one that
127 | is widely used among developers working in that language.
128 |
129 | The "System Libraries" of an executable work include anything, other
130 | than the work as a whole, that (a) is included in the normal form of
131 | packaging a Major Component, but which is not part of that Major
132 | Component, and (b) serves only to enable use of the work with that
133 | Major Component, or to implement a Standard Interface for which an
134 | implementation is available to the public in source code form. A
135 | "Major Component", in this context, means a major essential component
136 | (kernel, window system, and so on) of the specific operating system
137 | (if any) on which the executable work runs, or a compiler used to
138 | produce the work, or an object code interpreter used to run it.
139 |
140 | The "Corresponding Source" for a work in object code form means all
141 | the source code needed to generate, install, and (for an executable
142 | work) run the object code and to modify the work, including scripts to
143 | control those activities. However, it does not include the work's
144 | System Libraries, or general-purpose tools or generally available free
145 | programs which are used unmodified in performing those activities but
146 | which are not part of the work. For example, Corresponding Source
147 | includes interface definition files associated with source files for
148 | the work, and the source code for shared libraries and dynamically
149 | linked subprograms that the work is specifically designed to require,
150 | such as by intimate data communication or control flow between those
151 | subprograms and other parts of the work.
152 |
153 | The Corresponding Source need not include anything that users can
154 | regenerate automatically from other parts of the Corresponding Source.
155 |
156 | The Corresponding Source for a work in source code form is that same
157 | work.
158 |
159 | #### 2. Basic Permissions.
160 |
161 | All rights granted under this License are granted for the term of
162 | copyright on the Program, and are irrevocable provided the stated
163 | conditions are met. This License explicitly affirms your unlimited
164 | permission to run the unmodified Program. The output from running a
165 | covered work is covered by this License only if the output, given its
166 | content, constitutes a covered work. This License acknowledges your
167 | rights of fair use or other equivalent, as provided by copyright law.
168 |
169 | You may make, run and propagate covered works that you do not convey,
170 | without conditions so long as your license otherwise remains in force.
171 | You may convey covered works to others for the sole purpose of having
172 | them make modifications exclusively for you, or provide you with
173 | facilities for running those works, provided that you comply with the
174 | terms of this License in conveying all material for which you do not
175 | control copyright. Those thus making or running the covered works for
176 | you must do so exclusively on your behalf, under your direction and
177 | control, on terms that prohibit them from making any copies of your
178 | copyrighted material outside their relationship with you.
179 |
180 | Conveying under any other circumstances is permitted solely under the
181 | conditions stated below. Sublicensing is not allowed; section 10 makes
182 | it unnecessary.
183 |
184 | #### 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
185 |
186 | No covered work shall be deemed part of an effective technological
187 | measure under any applicable law fulfilling obligations under article
188 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or
189 | similar laws prohibiting or restricting circumvention of such
190 | measures.
191 |
192 | When you convey a covered work, you waive any legal power to forbid
193 | circumvention of technological measures to the extent such
194 | circumvention is effected by exercising rights under this License with
195 | respect to the covered work, and you disclaim any intention to limit
196 | operation or modification of the work as a means of enforcing, against
197 | the work's users, your or third parties' legal rights to forbid
198 | circumvention of technological measures.
199 |
200 | #### 4. Conveying Verbatim Copies.
201 |
202 | You may convey verbatim copies of the Program's source code as you
203 | receive it, in any medium, provided that you conspicuously and
204 | appropriately publish on each copy an appropriate copyright notice;
205 | keep intact all notices stating that this License and any
206 | non-permissive terms added in accord with section 7 apply to the code;
207 | keep intact all notices of the absence of any warranty; and give all
208 | recipients a copy of this License along with the Program.
209 |
210 | You may charge any price or no price for each copy that you convey,
211 | and you may offer support or warranty protection for a fee.
212 |
213 | #### 5. Conveying Modified Source Versions.
214 |
215 | You may convey a work based on the Program, or the modifications to
216 | produce it from the Program, in the form of source code under the
217 | terms of section 4, provided that you also meet all of these
218 | conditions:
219 |
220 | - a) The work must carry prominent notices stating that you modified
221 | it, and giving a relevant date.
222 | - b) The work must carry prominent notices stating that it is
223 | released under this License and any conditions added under
224 | section 7. This requirement modifies the requirement in section 4
225 | to "keep intact all notices".
226 | - c) You must license the entire work, as a whole, under this
227 | License to anyone who comes into possession of a copy. This
228 | License will therefore apply, along with any applicable section 7
229 | additional terms, to the whole of the work, and all its parts,
230 | regardless of how they are packaged. This License gives no
231 | permission to license the work in any other way, but it does not
232 | invalidate such permission if you have separately received it.
233 | - d) If the work has interactive user interfaces, each must display
234 | Appropriate Legal Notices; however, if the Program has interactive
235 | interfaces that do not display Appropriate Legal Notices, your
236 | work need not make them do so.
237 |
238 | A compilation of a covered work with other separate and independent
239 | works, which are not by their nature extensions of the covered work,
240 | and which are not combined with it such as to form a larger program,
241 | in or on a volume of a storage or distribution medium, is called an
242 | "aggregate" if the compilation and its resulting copyright are not
243 | used to limit the access or legal rights of the compilation's users
244 | beyond what the individual works permit. Inclusion of a covered work
245 | in an aggregate does not cause this License to apply to the other
246 | parts of the aggregate.
247 |
248 | #### 6. Conveying Non-Source Forms.
249 |
250 | You may convey a covered work in object code form under the terms of
251 | sections 4 and 5, provided that you also convey the machine-readable
252 | Corresponding Source under the terms of this License, in one of these
253 | ways:
254 |
255 | - a) Convey the object code in, or embodied in, a physical product
256 | (including a physical distribution medium), accompanied by the
257 | Corresponding Source fixed on a durable physical medium
258 | customarily used for software interchange.
259 | - b) Convey the object code in, or embodied in, a physical product
260 | (including a physical distribution medium), accompanied by a
261 | written offer, valid for at least three years and valid for as
262 | long as you offer spare parts or customer support for that product
263 | model, to give anyone who possesses the object code either (1) a
264 | copy of the Corresponding Source for all the software in the
265 | product that is covered by this License, on a durable physical
266 | medium customarily used for software interchange, for a price no
267 | more than your reasonable cost of physically performing this
268 | conveying of source, or (2) access to copy the Corresponding
269 | Source from a network server at no charge.
270 | - c) Convey individual copies of the object code with a copy of the
271 | written offer to provide the Corresponding Source. This
272 | alternative is allowed only occasionally and noncommercially, and
273 | only if you received the object code with such an offer, in accord
274 | with subsection 6b.
275 | - d) Convey the object code by offering access from a designated
276 | place (gratis or for a charge), and offer equivalent access to the
277 | Corresponding Source in the same way through the same place at no
278 | further charge. You need not require recipients to copy the
279 | Corresponding Source along with the object code. If the place to
280 | copy the object code is a network server, the Corresponding Source
281 | may be on a different server (operated by you or a third party)
282 | that supports equivalent copying facilities, provided you maintain
283 | clear directions next to the object code saying where to find the
284 | Corresponding Source. Regardless of what server hosts the
285 | Corresponding Source, you remain obligated to ensure that it is
286 | available for as long as needed to satisfy these requirements.
287 | - e) Convey the object code using peer-to-peer transmission,
288 | provided you inform other peers where the object code and
289 | Corresponding Source of the work are being offered to the general
290 | public at no charge under subsection 6d.
291 |
292 | A separable portion of the object code, whose source code is excluded
293 | from the Corresponding Source as a System Library, need not be
294 | included in conveying the object code work.
295 |
296 | A "User Product" is either (1) a "consumer product", which means any
297 | tangible personal property which is normally used for personal,
298 | family, or household purposes, or (2) anything designed or sold for
299 | incorporation into a dwelling. In determining whether a product is a
300 | consumer product, doubtful cases shall be resolved in favor of
301 | coverage. For a particular product received by a particular user,
302 | "normally used" refers to a typical or common use of that class of
303 | product, regardless of the status of the particular user or of the way
304 | in which the particular user actually uses, or expects or is expected
305 | to use, the product. A product is a consumer product regardless of
306 | whether the product has substantial commercial, industrial or
307 | non-consumer uses, unless such uses represent the only significant
308 | mode of use of the product.
309 |
310 | "Installation Information" for a User Product means any methods,
311 | procedures, authorization keys, or other information required to
312 | install and execute modified versions of a covered work in that User
313 | Product from a modified version of its Corresponding Source. The
314 | information must suffice to ensure that the continued functioning of
315 | the modified object code is in no case prevented or interfered with
316 | solely because modification has been made.
317 |
318 | If you convey an object code work under this section in, or with, or
319 | specifically for use in, a User Product, and the conveying occurs as
320 | part of a transaction in which the right of possession and use of the
321 | User Product is transferred to the recipient in perpetuity or for a
322 | fixed term (regardless of how the transaction is characterized), the
323 | Corresponding Source conveyed under this section must be accompanied
324 | by the Installation Information. But this requirement does not apply
325 | if neither you nor any third party retains the ability to install
326 | modified object code on the User Product (for example, the work has
327 | been installed in ROM).
328 |
329 | The requirement to provide Installation Information does not include a
330 | requirement to continue to provide support service, warranty, or
331 | updates for a work that has been modified or installed by the
332 | recipient, or for the User Product in which it has been modified or
333 | installed. Access to a network may be denied when the modification
334 | itself materially and adversely affects the operation of the network
335 | or violates the rules and protocols for communication across the
336 | network.
337 |
338 | Corresponding Source conveyed, and Installation Information provided,
339 | in accord with this section must be in a format that is publicly
340 | documented (and with an implementation available to the public in
341 | source code form), and must require no special password or key for
342 | unpacking, reading or copying.
343 |
344 | #### 7. Additional Terms.
345 |
346 | "Additional permissions" are terms that supplement the terms of this
347 | License by making exceptions from one or more of its conditions.
348 | Additional permissions that are applicable to the entire Program shall
349 | be treated as though they were included in this License, to the extent
350 | that they are valid under applicable law. If additional permissions
351 | apply only to part of the Program, that part may be used separately
352 | under those permissions, but the entire Program remains governed by
353 | this License without regard to the additional permissions.
354 |
355 | When you convey a copy of a covered work, you may at your option
356 | remove any additional permissions from that copy, or from any part of
357 | it. (Additional permissions may be written to require their own
358 | removal in certain cases when you modify the work.) You may place
359 | additional permissions on material, added by you to a covered work,
360 | for which you have or can give appropriate copyright permission.
361 |
362 | Notwithstanding any other provision of this License, for material you
363 | add to a covered work, you may (if authorized by the copyright holders
364 | of that material) supplement the terms of this License with terms:
365 |
366 | - a) Disclaiming warranty or limiting liability differently from the
367 | terms of sections 15 and 16 of this License; or
368 | - b) Requiring preservation of specified reasonable legal notices or
369 | author attributions in that material or in the Appropriate Legal
370 | Notices displayed by works containing it; or
371 | - c) Prohibiting misrepresentation of the origin of that material,
372 | or requiring that modified versions of such material be marked in
373 | reasonable ways as different from the original version; or
374 | - d) Limiting the use for publicity purposes of names of licensors
375 | or authors of the material; or
376 | - e) Declining to grant rights under trademark law for use of some
377 | trade names, trademarks, or service marks; or
378 | - f) Requiring indemnification of licensors and authors of that
379 | material by anyone who conveys the material (or modified versions
380 | of it) with contractual assumptions of liability to the recipient,
381 | for any liability that these contractual assumptions directly
382 | impose on those licensors and authors.
383 |
384 | All other non-permissive additional terms are considered "further
385 | restrictions" within the meaning of section 10. If the Program as you
386 | received it, or any part of it, contains a notice stating that it is
387 | governed by this License along with a term that is a further
388 | restriction, you may remove that term. If a license document contains
389 | a further restriction but permits relicensing or conveying under this
390 | License, you may add to a covered work material governed by the terms
391 | of that license document, provided that the further restriction does
392 | not survive such relicensing or conveying.
393 |
394 | If you add terms to a covered work in accord with this section, you
395 | must place, in the relevant source files, a statement of the
396 | additional terms that apply to those files, or a notice indicating
397 | where to find the applicable terms.
398 |
399 | Additional terms, permissive or non-permissive, may be stated in the
400 | form of a separately written license, or stated as exceptions; the
401 | above requirements apply either way.
402 |
403 | #### 8. Termination.
404 |
405 | You may not propagate or modify a covered work except as expressly
406 | provided under this License. Any attempt otherwise to propagate or
407 | modify it is void, and will automatically terminate your rights under
408 | this License (including any patent licenses granted under the third
409 | paragraph of section 11).
410 |
411 | However, if you cease all violation of this License, then your license
412 | from a particular copyright holder is reinstated (a) provisionally,
413 | unless and until the copyright holder explicitly and finally
414 | terminates your license, and (b) permanently, if the copyright holder
415 | fails to notify you of the violation by some reasonable means prior to
416 | 60 days after the cessation.
417 |
418 | Moreover, your license from a particular copyright holder is
419 | reinstated permanently if the copyright holder notifies you of the
420 | violation by some reasonable means, this is the first time you have
421 | received notice of violation of this License (for any work) from that
422 | copyright holder, and you cure the violation prior to 30 days after
423 | your receipt of the notice.
424 |
425 | Termination of your rights under this section does not terminate the
426 | licenses of parties who have received copies or rights from you under
427 | this License. If your rights have been terminated and not permanently
428 | reinstated, you do not qualify to receive new licenses for the same
429 | material under section 10.
430 |
431 | #### 9. Acceptance Not Required for Having Copies.
432 |
433 | You are not required to accept this License in order to receive or run
434 | a copy of the Program. Ancillary propagation of a covered work
435 | occurring solely as a consequence of using peer-to-peer transmission
436 | to receive a copy likewise does not require acceptance. However,
437 | nothing other than this License grants you permission to propagate or
438 | modify any covered work. These actions infringe copyright if you do
439 | not accept this License. Therefore, by modifying or propagating a
440 | covered work, you indicate your acceptance of this License to do so.
441 |
442 | #### 10. Automatic Licensing of Downstream Recipients.
443 |
444 | Each time you convey a covered work, the recipient automatically
445 | receives a license from the original licensors, to run, modify and
446 | propagate that work, subject to this License. You are not responsible
447 | for enforcing compliance by third parties with this License.
448 |
449 | An "entity transaction" is a transaction transferring control of an
450 | organization, or substantially all assets of one, or subdividing an
451 | organization, or merging organizations. If propagation of a covered
452 | work results from an entity transaction, each party to that
453 | transaction who receives a copy of the work also receives whatever
454 | licenses to the work the party's predecessor in interest had or could
455 | give under the previous paragraph, plus a right to possession of the
456 | Corresponding Source of the work from the predecessor in interest, if
457 | the predecessor has it or can get it with reasonable efforts.
458 |
459 | You may not impose any further restrictions on the exercise of the
460 | rights granted or affirmed under this License. For example, you may
461 | not impose a license fee, royalty, or other charge for exercise of
462 | rights granted under this License, and you may not initiate litigation
463 | (including a cross-claim or counterclaim in a lawsuit) alleging that
464 | any patent claim is infringed by making, using, selling, offering for
465 | sale, or importing the Program or any portion of it.
466 |
467 | #### 11. Patents.
468 |
469 | A "contributor" is a copyright holder who authorizes use under this
470 | License of the Program or a work on which the Program is based. The
471 | work thus licensed is called the contributor's "contributor version".
472 |
473 | A contributor's "essential patent claims" are all patent claims owned
474 | or controlled by the contributor, whether already acquired or
475 | hereafter acquired, that would be infringed by some manner, permitted
476 | by this License, of making, using, or selling its contributor version,
477 | but do not include claims that would be infringed only as a
478 | consequence of further modification of the contributor version. For
479 | purposes of this definition, "control" includes the right to grant
480 | patent sublicenses in a manner consistent with the requirements of
481 | this License.
482 |
483 | Each contributor grants you a non-exclusive, worldwide, royalty-free
484 | patent license under the contributor's essential patent claims, to
485 | make, use, sell, offer for sale, import and otherwise run, modify and
486 | propagate the contents of its contributor version.
487 |
488 | In the following three paragraphs, a "patent license" is any express
489 | agreement or commitment, however denominated, not to enforce a patent
490 | (such as an express permission to practice a patent or covenant not to
491 | sue for patent infringement). To "grant" such a patent license to a
492 | party means to make such an agreement or commitment not to enforce a
493 | patent against the party.
494 |
495 | If you convey a covered work, knowingly relying on a patent license,
496 | and the Corresponding Source of the work is not available for anyone
497 | to copy, free of charge and under the terms of this License, through a
498 | publicly available network server or other readily accessible means,
499 | then you must either (1) cause the Corresponding Source to be so
500 | available, or (2) arrange to deprive yourself of the benefit of the
501 | patent license for this particular work, or (3) arrange, in a manner
502 | consistent with the requirements of this License, to extend the patent
503 | license to downstream recipients. "Knowingly relying" means you have
504 | actual knowledge that, but for the patent license, your conveying the
505 | covered work in a country, or your recipient's use of the covered work
506 | in a country, would infringe one or more identifiable patents in that
507 | country that you have reason to believe are valid.
508 |
509 | If, pursuant to or in connection with a single transaction or
510 | arrangement, you convey, or propagate by procuring conveyance of, a
511 | covered work, and grant a patent license to some of the parties
512 | receiving the covered work authorizing them to use, propagate, modify
513 | or convey a specific copy of the covered work, then the patent license
514 | you grant is automatically extended to all recipients of the covered
515 | work and works based on it.
516 |
517 | A patent license is "discriminatory" if it does not include within the
518 | scope of its coverage, prohibits the exercise of, or is conditioned on
519 | the non-exercise of one or more of the rights that are specifically
520 | granted under this License. You may not convey a covered work if you
521 | are a party to an arrangement with a third party that is in the
522 | business of distributing software, under which you make payment to the
523 | third party based on the extent of your activity of conveying the
524 | work, and under which the third party grants, to any of the parties
525 | who would receive the covered work from you, a discriminatory patent
526 | license (a) in connection with copies of the covered work conveyed by
527 | you (or copies made from those copies), or (b) primarily for and in
528 | connection with specific products or compilations that contain the
529 | covered work, unless you entered into that arrangement, or that patent
530 | license was granted, prior to 28 March 2007.
531 |
532 | Nothing in this License shall be construed as excluding or limiting
533 | any implied license or other defenses to infringement that may
534 | otherwise be available to you under applicable patent law.
535 |
536 | #### 12. No Surrender of Others' Freedom.
537 |
538 | If conditions are imposed on you (whether by court order, agreement or
539 | otherwise) that contradict the conditions of this License, they do not
540 | excuse you from the conditions of this License. If you cannot convey a
541 | covered work so as to satisfy simultaneously your obligations under
542 | this License and any other pertinent obligations, then as a
543 | consequence you may not convey it at all. For example, if you agree to
544 | terms that obligate you to collect a royalty for further conveying
545 | from those to whom you convey the Program, the only way you could
546 | satisfy both those terms and this License would be to refrain entirely
547 | from conveying the Program.
548 |
549 | #### 13. Use with the GNU Affero General Public License.
550 |
551 | Notwithstanding any other provision of this License, you have
552 | permission to link or combine any covered work with a work licensed
553 | under version 3 of the GNU Affero General Public License into a single
554 | combined work, and to convey the resulting work. The terms of this
555 | License will continue to apply to the part which is the covered work,
556 | but the special requirements of the GNU Affero General Public License,
557 | section 13, concerning interaction through a network will apply to the
558 | combination as such.
559 |
560 | #### 14. Revised Versions of this License.
561 |
562 | The Free Software Foundation may publish revised and/or new versions
563 | of the GNU General Public License from time to time. Such new versions
564 | will be similar in spirit to the present version, but may differ in
565 | detail to address new problems or concerns.
566 |
567 | Each version is given a distinguishing version number. If the Program
568 | specifies that a certain numbered version of the GNU General Public
569 | License "or any later version" applies to it, you have the option of
570 | following the terms and conditions either of that numbered version or
571 | of any later version published by the Free Software Foundation. If the
572 | Program does not specify a version number of the GNU General Public
573 | License, you may choose any version ever published by the Free
574 | Software Foundation.
575 |
576 | If the Program specifies that a proxy can decide which future versions
577 | of the GNU General Public License can be used, that proxy's public
578 | statement of acceptance of a version permanently authorizes you to
579 | choose that version for the Program.
580 |
581 | Later license versions may give you additional or different
582 | permissions. However, no additional obligations are imposed on any
583 | author or copyright holder as a result of your choosing to follow a
584 | later version.
585 |
586 | #### 15. Disclaimer of Warranty.
587 |
588 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
589 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
590 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT
591 | WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT
592 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
593 | A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND
594 | PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE
595 | DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR
596 | CORRECTION.
597 |
598 | #### 16. Limitation of Liability.
599 |
600 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
601 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR
602 | CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
603 | INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES
604 | ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT
605 | NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR
606 | LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM
607 | TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER
608 | PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
609 |
610 | #### 17. Interpretation of Sections 15 and 16.
611 |
612 | If the disclaimer of warranty and limitation of liability provided
613 | above cannot be given local legal effect according to their terms,
614 | reviewing courts shall apply local law that most closely approximates
615 | an absolute waiver of all civil liability in connection with the
616 | Program, unless a warranty or assumption of liability accompanies a
617 | copy of the Program in return for a fee.
618 |
619 | END OF TERMS AND CONDITIONS
620 |
621 | ### How to Apply These Terms to Your New Programs
622 |
623 | If you develop a new program, and you want it to be of the greatest
624 | possible use to the public, the best way to achieve this is to make it
625 | free software which everyone can redistribute and change under these
626 | terms.
627 |
628 | To do so, attach the following notices to the program. It is safest to
629 | attach them to the start of each source file to most effectively state
630 | the exclusion of warranty; and each file should have at least the
631 | "copyright" line and a pointer to where the full notice is found.
632 |
633 |
634 | Copyright (C)
635 |
636 | This program is free software: you can redistribute it and/or modify
637 | it under the terms of the GNU General Public License as published by
638 | the Free Software Foundation, either version 3 of the License, or
639 | (at your option) any later version.
640 |
641 | This program is distributed in the hope that it will be useful,
642 | but WITHOUT ANY WARRANTY; without even the implied warranty of
643 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
644 | GNU General Public License for more details.
645 |
646 | You should have received a copy of the GNU General Public License
647 | along with this program. If not, see .
648 |
649 | Also add information on how to contact you by electronic and paper
650 | mail.
651 |
652 | If the program does terminal interaction, make it output a short
653 | notice like this when it starts in an interactive mode:
654 |
655 | Copyright (C)
656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
657 | This is free software, and you are welcome to redistribute it
658 | under certain conditions; type `show c' for details.
659 |
660 | The hypothetical commands \`show w' and \`show c' should show the
661 | appropriate parts of the General Public License. Of course, your
662 | program's commands might be different; for a GUI interface, you would
663 | use an "about box".
664 |
665 | You should also get your employer (if you work as a programmer) or
666 | school, if any, to sign a "copyright disclaimer" for the program, if
667 | necessary. For more information on this, and how to apply and follow
668 | the GNU GPL, see .
669 |
670 | The GNU General Public License does not permit incorporating your
671 | program into proprietary programs. If your program is a subroutine
672 | library, you may consider it more useful to permit linking proprietary
673 | applications with the library. If this is what you want to do, use the
674 | GNU Lesser General Public License instead of this License. But first,
675 | please read .
676 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Toggle_Switch/Makefile:
--------------------------------------------------------------------------------
1 |
2 | all:
3 | make toggle_switch_ABC_pps_vec_par
4 | make toggle_switch_ABC_pps_novec_nopar
5 | make toggle_switch_ABC_pps_novec_par
6 |
7 | toggle_switch_ABC_pps_vec_par: toggle_switch_ABC_pps_vec_par.c Makefile
8 | icc -mkl=sequential -qopenmp -O2 -xHost -fcf-protection=none toggle_switch_ABC_pps_vec_par.c -o toggle_switch_ABC_pps_vec_par
9 |
10 | toggle_switch_ABC_pps_novec_par: toggle_switch_ABC_pps_vec_par.c Makefile
11 | icc -mkl=sequential -qopenmp -O2 -xHost -qno-openmp-simd -no-vec toggle_switch_ABC_pps_vec_par.c -o toggle_switch_ABC_pps_novec_par
12 |
13 | toggle_switch_ABC_pps_novec_nopar: toggle_switch_ABC_pps_vec_par.c Makefile
14 | icc -mkl=sequential -qno-openmp -O2 -xHost -no-vec toggle_switch_ABC_pps_novec_nopar.c -o toggle_switch_ABC_pps_novec_nopar
15 |
16 | clean:
17 | rm -f toggle_switch_ABC_pps_vec_par toggle_switch_ABC_pps_novec_nopar toggle_switch_ABC_pps_novec_par
18 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Toggle_Switch/run_bench.sh:
--------------------------------------------------------------------------------
1 | #!/bin/bash
2 |
3 | N=4000
4 | C=8000
5 | T=600
6 | SEED=1337
7 |
8 | echo "Toggle switch prior predicive sampling benchmark"
9 | echo "N = $N C = $C T = $T"
10 |
11 | #run and time vectorised parallel version
12 | echo 'timing parallel+SIMD'
13 | time ./toggle_switch_ABC_pps_vec_par $N $SEED $C $T > /dev/null
14 | echo 'done'
15 |
16 | #run and time scalar parallel version
17 | # (still uses OpenMP and MKL/VSL, but SIMD and auto-vectorisation disabled)
18 | echo 'timing parallel+scalar'
19 | time ./toggle_switch_ABC_pps_novec_par $N $SEED $C $T > /dev/null
20 | echo 'done'
21 |
22 | #run and time scalar sequential version
23 | # (still uses MKL/VSL, but OpenMP and auto-vectorisation disabled )
24 | echo 'timing sequential+scalar'
25 | time ./toggle_switch_ABC_pps_novec_nopar $N $SEED $C $T > /dev/null
26 | echo 'done'
27 |
28 |
29 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Toggle_Switch/toggle_switch_ABC_pps_naive.R:
--------------------------------------------------------------------------------
1 | library(tictoc)
2 | library(doParallel)
3 |
4 | # Naive implementation of toggle switch simulation
5 | simulate_toggle_switch <- function(mu,sigma,gam, alpha, beta, T, C) {
6 |
7 | # create ouput array
8 | y <- numeric(C)
9 | # evolve each cell one-by-one
10 | for (i in 1:C) {
11 | # initialise
12 | alpha_u <- alpha[1]
13 | alpha_v <- alpha[2]
14 | beta_u <- beta[1]
15 | beta_v <- beta[2]
16 | u_t <- 10
17 | v_t <- 10
18 | for (j in 2:T){
19 | p_u <- v_t^beta_u
20 | p_v <- u_t^beta_v
21 | u_t <- 0.97*u_t + alpha_u/(1+p_u) - 1.0 + 0.5*rnorm(1,0,1)
22 | v_t <- 0.97*v_t + alpha_v/(1+p_v) - 1.0 + 0.5*rnorm(1,0,1)
23 | if (u_t < 1.0 ){
24 | u_t <- 1.0
25 | }
26 | if (v_t < 1.0) {
27 | v_t <- 1.0
28 | }
29 | }
30 | y[i] <- u_t + sigma*mu*rnorm(1,0,1)/(u_t^gam)
31 | if (y[i] < 1.0) {
32 | y <- 1.0
33 | }
34 | }
35 | return(y)
36 | }
37 |
38 | # prior predictive sampling
39 | T <- 600
40 | C <- 8000
41 | N <- 240
42 |
43 | for (P in cores) {
44 | # set up level of parallelism
45 | cl <- makeCluster(P)
46 | registerDoParallel(cl)
47 | # run optimised R simulations
48 | tic()
49 | obs_vals2 <- foreach(k = 1:N) %dopar% {
50 | theta <- runif(7,c(250.0,0.05,0.05,0.0,0.0,0.0,0.0),
51 | c(400.0,0.5,0.35,50.0,50.0,7.0,7.0))
52 | mu <- theta[1]
53 | sigma <- theta[2]
54 | gam <- theta[3]
55 | alpha <- theta[4:5]
56 | beta <- theta[6:7]
57 |
58 | c(theta,simulate_toggle_switch(mu,sigma,gam,alpha,beta,T,C))
59 | }
60 | print(c(P,N,C,T))
61 | toc()
62 | stopCluster(cl)
63 | }
64 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Toggle_Switch/toggle_switch_ABC_pps_opt.R:
--------------------------------------------------------------------------------
1 | library(tictoc)
2 | library(doParallel)
3 |
4 | # optimized R-base implementation of the toggle switch model
5 | simulate_toggle_switch_vec <- function(mu,sigma,gam, alpha, beta, T, C) {
6 |
7 | # create ouput array
8 | u_t <- numeric(C)
9 | v_t <- numeric(C)
10 |
11 | # initialise
12 | alpha_u <- alpha[1]
13 | alpha_v <- alpha[2]
14 | beta_u <- beta[1]
15 | beta_v <- beta[2]
16 | u_t[1:C] <- 10
17 | v_t[1:C] <- 10
18 | # generate random variates
19 | zeta <- matrix(nrow=C,ncol=2*(T-1)+1)
20 | zeta[,] <- rnorm(C*(2*(T-1)+1),0,1)
21 | # evolve all cells together
22 | for (j in 2:T) {
23 | p_u <- v_t^beta_u
24 | p_v <- u_t^beta_v
25 | u_t <- 0.97*u_t + alpha_u/(1+p_u) - 1.0 + 0.5*zeta[1:C,2*(j-1)]
26 | v_t <- 0.97*v_t + alpha_v/(1+p_v) - 1.0 + 0.5*zeta[1:C,2*(j-1) + 1]
27 | u_t[u_t < 1.0] <- 1.0;
28 | v_t[v_t < 1.0] <- 1.0;
29 |
30 | }
31 | y <- u_t + sigma*mu*zeta[1:C,1]/(u_t^gam)
32 | y[y < 1.0] <- 1.0
33 | return(y)
34 | }
35 |
36 | # prior predictive sampling
37 | T <- 600
38 | C <- 8000
39 | N <- 8064
40 |
41 | for (P in cores) {
42 | # set up level of parallelism
43 | cl <- makeCluster(P)
44 | registerDoParallel(cl)
45 | # run optimised R simulations
46 | tic()
47 | obs_vals2 <- foreach(k = 1:N) %dopar% {
48 | theta <- runif(7,c(250.0,0.05,0.05,0.0,0.0,0.0,0.0),
49 | c(400.0,0.5,0.35,50.0,50.0,7.0,7.0))
50 | mu <- theta[1]
51 | sigma <- theta[2]
52 | gam <- theta[3]
53 | alpha <- theta[4:5]
54 | beta <- theta[6:7]
55 |
56 | c(theta,simulate_toggle_switch_vec(mu,sigma,gam,alpha,beta,T,C))
57 | }
58 | print(c(P,N,C,T))
59 | toc()
60 | stopCluster(cl)
61 | }
62 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Toggle_Switch/toggle_switch_ABC_pps_vec_par.c:
--------------------------------------------------------------------------------
1 | /* Bayesian computations using SIMD operations
2 | * Copyright (C) 2019 David J. Warne, Christopher C. Drovandi
3 | *
4 | * This program is free software: you can redistribute it and/or modify
5 | * it under the terms of the GNU General Public License as published by
6 | * the Free Software Foundation, either version 3 of the License, or
7 | * (at your option) any later version.
8 | *
9 | * This program is distributed in the hope that it will be useful,
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 | * GNU General Public License for more details.
13 | *
14 | * You should have received a copy of the GNU General Public License
15 | * along with this program. If not, see .
16 | */
17 | /**
18 | * @file toggle_switch_ABC_pps.c
19 | *
20 | * @brief Demonstration of vectorisation for approximate Bayesian computation
21 | *
22 | * @details Efficient prior predictive sampling for the genetic toggle switch
23 | * model. Cells are evoluted in groups of equal to the vector processing unit
24 | * capacity for double precision floating point. Standard Gaussian random
25 | * variates are generated for the entire evolution of the group to efficiently
26 | * utilise the MKL VSL generator.
27 | *
28 | * @author David J. Warne (david.warne@qut.edu.au)
29 | * School of Mathematical Sciences
30 | * Queensland University of Technology
31 | *
32 | * @author Chris C. Drovandi (c.drovandi@qut.edu.au)
33 | * School of Mathematical Sciences
34 | * Queensland University of Technology
35 | *
36 | * @date 12 Oct 2018
37 | *
38 | */
39 |
40 | /* standard C headers*/
41 | #include
42 | #include
43 | #include
44 |
45 | /* Intel headers */
46 | #include
47 | #include
48 |
49 | /* OpenMP header */
50 | #include
51 |
52 | /* length of vector processing units and ideal memory alignement*/
53 | #if defined(__AVX512BW__)
54 | #define VECLEN 8
55 | #define ALIGN 64
56 | #elif defined(__AVX2__)
57 | #define VECLEN 4
58 | #define ALIGN 64
59 | #elif defined(__AVX__)
60 | #define VECLEN 4
61 | #define ALIGN 32
62 | #elif defined(__SSE4_2__)
63 | #define VECLEN 2
64 | #define ALIGN 16
65 | #endif
66 |
67 | /**
68 | * @brief vectorised toggle switch stochastic simulation
69 | *
70 | * @details processes cells in blocks of VECLEN to exploit vectorisation and
71 | * cache. Utilises MKL VSL routine to generate all Gaussian random variates for
72 | * the entire evolution of the block.
73 | *
74 | * @param stream Pointer to RNG state
75 | * @param mu basal observation noise level
76 | * @param sigma,gamma increase the observation noise at low expression levels
77 | * @param alpha, beta logistic-like repression of gene expression
78 | * @param T simulation time and observation time
79 | * @param C number of cells
80 | * @param zeta memory location to store gaussian random variaates
81 | * @param y vector of C observations of u-gene expression levels
82 | *
83 | * @note assumes argument data arrays are aligned on ALIGN-byte boundary.
84 | * The restrict keyword is used to ensure the compile knows there is no pointer
85 | * aliasing.
86 | *
87 | * @warning For optimal performance, the number of cells, C, should be a
88 | * multiple of VECLEN
89 | */
90 | void
91 | simulate_toggle_switch(VSLStreamStatePtr stream,
92 | double mu, double sigma, double gamma,
93 | double * restrict alpha, double * restrict beta,
94 | int T, int C,double * restrict zeta, double * restrict y)
95 | {
96 |
97 | /* process cells in blocks of VECLEN*/
98 | for (int c=0;c= 1.0) ? u_t : 1.0;
136 | v_t += 0.5*zeta_v;
137 | v_t = (v_t >= 1.0) ? v_t : 1.0;
138 | }
139 |
140 | /* make noisy observation */
141 | y[c+c2] = u_t + _mu + _sigma*_mu*zeta[(T-1)*VECLEN*2 + c2]/pow(u_t,_gamma);
142 | y[c+c2] = (y[c+c2] >= 1.0) ? y[c+c2]: 1.0;
143 | }
144 | }
145 | }
146 |
147 | /**
148 | * @brief program entry point
149 | * @details Generates prior predictive samples for the genetic toggle switch
150 | * model. Distributes simulations across availablle cores and utilises fine grain
151 | * SIMD operations for groups of cells within each simulation.
152 | *
153 | * @param argc number of command line arguments
154 | * @param argv vector of argument strings
155 | */
156 | int
157 | main(int argc,char **argv)
158 | {
159 | int T, C, K;
160 | double *theta, *obs_vals;
161 | int seed, sims;
162 |
163 | /* default number of simulation timesteps*/
164 | T = 300;
165 |
166 | /* default number of cells*/
167 | C = 2000;
168 |
169 | /* number of parameters*/
170 | K = 7;
171 |
172 | /* get command line arguments*/
173 | if (argc < 2)
174 | {
175 | fprintf(stderr,"Usage : [%s] sims seed [C] [T] [VEC %d ALIGN %d\n",argv[0],VECLEN,ALIGN);
176 | exit(1);
177 | }
178 | sims = (int)atoi(argv[1]);
179 | seed = (int)atoi(argv[2]);
180 |
181 | /* check if C was specified*/
182 | if (argc > 3)
183 | {
184 | C = (int)atoi(argv[3]);
185 | if (C%VECLEN != 0)
186 | {
187 | fprintf(stderr,"Warning: For optimal performance C must be a multiple of %d\n",VECLEN);
188 | }
189 | }
190 | /* check if T was specified*/
191 | if (argc > 4)
192 | {
193 | T = (int)atoi(argv[4]);
194 | }
195 |
196 | /*allocate aligned memory for noisy observations*/
197 | obs_vals = (double *)_mm_malloc(C*sims*sizeof(double),ALIGN);
198 |
199 | /* allocate aligned memory prior samples that generated the data*/
200 | theta = (double *)_mm_malloc(K*sims*sizeof(double),ALIGN);
201 |
202 | /* compute simulations in parallel*/
203 | #pragma omp parallel shared(seed,sims,C,obs_vals,theta)
204 | {
205 | VSLStreamStatePtr stream;
206 | int thread_id, num_threads, sims_per_thread;
207 | double *zeta;
208 |
209 | /* get thread information and assign workload*/
210 | thread_id = omp_get_thread_num();
211 | num_threads = omp_get_num_threads();
212 | sims_per_thread = sims/num_threads;
213 |
214 | fprintf(stderr,"ID: %d, total: %d sims : %d\n",thread_id,num_threads, sims_per_thread);
215 | /* initialise RNG stream for this thread*/
216 | vslNewStream(&stream,VSL_BRNG_MT2203+thread_id,seed);
217 |
218 | /*allocate aligned memory for Gaussian random variates*/
219 | zeta = (double *)_mm_malloc(2*VECLEN*T*sizeof(double),ALIGN);
220 |
221 | /* compute simulations in this threads workload*/
222 | for (int k=thread_id*sims_per_thread;k<(thread_id+1)*sims_per_thread;k++)
223 | {
224 | /* model parameters */
225 | double mu,sigma,gamma; /* measurement error parameters */
226 | double alpha[2], beta[2]; /* parameters for gene expression*/
227 |
228 | /*sample prior */
229 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,1,&mu,250.0,400.0);
230 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,1,&sigma,0.05,0.5);
231 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,1,&gamma,0.05,0.35);
232 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,2,alpha,0.0,50.0);
233 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,2,beta,0.0,7.0);
234 |
235 | /*run simulation and store observations*/
236 | simulate_toggle_switch(stream,mu,sigma,gamma,alpha,beta,T,C,zeta,
237 | obs_vals +k*C);
238 | /*store prior sample*/
239 | theta[k*7] = mu;
240 | theta[k*7 + 1] = sigma;
241 | theta[k*7 + 2] = gamma;
242 | theta[k*7 + 3] = alpha[0];
243 | theta[k*7 + 4] = beta[0];
244 | theta[k*7 + 5] = alpha[1];
245 | theta[k*7 + 6] = beta[1];
246 | }
247 | /*clean up memory*/
248 | vslDeleteStream(&stream);
249 | _mm_free(zeta);
250 | }
251 |
252 | /*output prior predicitive samples for postprocessing for ABC*/
253 | fprintf(stdout,"\"Sample\",\"mu\",\"sigma\",\"gamma\",\"alpha_u\",\"beta_u\",\"alpha_v\",\"beta_v\"");
254 | for (int j=0;j /dev/null
12 | echo 'done'
13 |
14 | #run and time scalar parallel version
15 | # (still uses OpenMP and MKL/VSL, but SIMD and auto-vectorisation disabled)
16 | echo 'timing parallel+scalar'
17 | time ./tuberculosis_ABC_pps_novec_par $N $SEED > /dev/null
18 | echo 'done'
19 |
20 | #run and time scalar sequential version
21 | # (still uses MKL/VSL, but OpenMP and auto-vectorisation disabled )
22 | echo 'timing sequential+scalar'
23 | time ./tuberculosis_ABC_pps_novec_nopar $N $SEED > /dev/null
24 | echo 'done'
25 |
26 |
--------------------------------------------------------------------------------
/Prior_Predictive_Sampling_Tuberculosis/tuberculosis_ABC_pps_vec_par.c:
--------------------------------------------------------------------------------
1 | /* Bayesian computations using SIMD operations
2 | * Copyright (C) 2019 David J. Warne, Christopher C. Drovandi, Scott A. Sisson
3 | *
4 | * This program is free software: you can redistribute it and/or modify
5 | * it under the terms of the GNU General Public License as published by
6 | * the Free Software Foundation, either version 3 of the License, or
7 | * (at your option) any later version.
8 | *
9 | * This program is distributed in the hope that it will be useful,
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 | * GNU General Public License for more details.
13 | *
14 | * You should have received a copy of the GNU General Public License
15 | * along with this program. If not, see .
16 | */
17 | /**
18 | * @file tuderculosis_ABC_pps.c
19 | * @brief Demonstration of vectorisation for approximate Bayesian computation
20 | * @details Uses the Disease transmission and mutation model
21 | *
22 | * @author David J. Warne (david.warne@qut.edu.au)
23 | * School of Mathematical Sciences
24 | * Queensland University of Technology
25 | *
26 | * @author Chris C. Drovandi (c.drovandi@qut.edu.au)
27 | * School of Mathematical Sciences
28 | * Queensland University of Technology
29 | *
30 | * @author Scott A. Sisson (scott.sisson@unsw.edu.au)
31 | * School of Mathematics and Statistics
32 | * University of New South Whales
33 | *
34 | * @date 1 Nov 2018
35 | *
36 | */
37 |
38 | /* standard C headers */
39 | #include
40 | #include
41 | #include
42 |
43 | /* Intel headers */
44 | #include
45 | #include
46 |
47 | /* OpenMP header */
48 | #include
49 |
50 | /* length of vector processing units and ideal memory alignment*/
51 | #if defined(__AVX512BW__)
52 | #define VECLEN 8
53 | #define ALIGN 64
54 | #elif defined(__AVX2__)
55 | #define VECLEN 4
56 | #define ALIGN 64
57 | #elif defined(__AVX__)
58 | #define VECLEN 4
59 | #define ALIGN 32
60 | #elif defined(__SSE4_2__)
61 | #define VECLEN 2
62 | #define ALIGN 16
63 | #endif
64 |
65 | #define BIRTH 0
66 | #define DEATH 1
67 | #define MUTATION 2
68 |
69 | #define MAXN 10000
70 | #define MAXG 1000
71 |
72 | /**
73 | * @brief discrete random variable sampler using lookup method
74 | *
75 | * @param stream Pointer to RNG state
76 | * @param p array of probabilities
77 | * @param n number of possible outcomes
78 | * @param ix pointer to store the selected outcome
79 | */
80 | void
81 | sample(VSLStreamStatePtr stream,double * restrict p,int n, int *ix)
82 | {
83 | double sum,u;
84 | int k;
85 |
86 | k = 0;
87 | sum = p[k];
88 | /* generate a uniform random variable and see where it lies in the cumulative prob */
89 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,1,&u, 0.0,1.0);
90 | while (u > sum && k < n-1)
91 | {
92 | k++;
93 | sum += p[k];
94 | }
95 | *ix = k;
96 | return;
97 | }
98 |
99 | /**
100 | * @brief partly vectorised TB dynamics stochastic simulation
101 | *
102 | * @details uses Gillespies direct method with some update operations being
103 | * vectorised. Example of an algorithm that is diffisult to vectorise without
104 | * introduction of approximations.
105 | *
106 | * @param stream Pointer to RNG state
107 | * @param alpha birth rate
108 | * @param delta death rate
109 | * @param mutation rate
110 | * @param g_ret pointer to store distinct genotype count
111 | * @param H_ret pointer to store genetic diveristy
112 |
113 | * @note assumes argument data arrays are aligned on ALIGN-byte boundary.
114 | * The restrict keyword is used to ensure the compile knows there is no pointer
115 | * aliasing.
116 | */
117 | void
118 | simulate(VSLStreamStatePtr stream, double alpha, double delta, double mu,
119 | double * restrict g_ret, double * restrict H_ret)
120 | {
121 |
122 | /* the current number of genotypes and population*/
123 | int G=1, N=1;
124 |
125 | /*maximum population*/
126 | int Nstop = MAXN;
127 |
128 | /*maximum genotypes*/
129 | int maxG = MAXN;
130 |
131 | /*birth/death/mutation probabilities*/
132 | __declspec(align(ALIGN)) double probs[3];
133 | double sumProbs;
134 |
135 | /*genotype probabilities*/
136 | __declspec(align(ALIGN)) double probs_G[MAXN];
137 |
138 | /* number of individuals in each genotype in population */
139 | __declspec(align(ALIGN)) double X[MAXN];
140 |
141 | /* number of individuals in each genotype in sample */
142 | __declspec(align(ALIGN)) double x[MAXN];
143 | double g,H;
144 | int event_val = 0, event_val_G = 0;
145 |
146 | /* initialise the numbers in each genotype */
147 | #pragma omp simd
148 | for (int i=0; i 0 && N < Nstop && G < maxG)
165 | {
166 | float Nf;
167 | Nf = (double)N;
168 |
169 | /* which event occurs?*/
170 | sample(stream,probs,3,&event_val);
171 |
172 | #pragma omp simd
173 | for (int i=0; i 0)
230 | {
231 | int numSamples = 473;
232 | int index;
233 | /* take a sample from the population */
234 | for (int i = 0; i 0)
240 | {
241 | /* remove sample from population (sampling without replacement) */
242 | X[index] = X[index]-1;
243 | /* add invidual into the sample */
244 | x[index] = x[index]+1;
245 | break;
246 | }
247 | }
248 | }
249 |
250 |
251 | /* compute the distinct genotypes */
252 | g = 0;
253 | #pragma omp simd reduction(+:g)
254 | for (int i=0; i 0);
257 | }
258 |
259 | /* compute the genetic diversity */
260 | H = 0;
261 | #pragma omp simd reduction(+:H)
262 | for (int i=0;i 0)
343 | {
344 | break;
345 | }
346 |
347 | }
348 | /* simulate pseudo-data*/
349 | simulate(stream,alpha, delta, mu, &g_ret, &H_ret);
350 |
351 | /*store prior samples*/
352 | theta[k*3] = alpha;
353 | theta[k*3 + 1] = delta;
354 | theta[k*3 + 2] = mu;
355 |
356 | /*store summary stats*/
357 | S[k*2] = g_ret;
358 | S[k*2 +1] = H_ret;
359 | }
360 | /*clean up memory*/
361 | vslDeleteStream(&stream);
362 | }
363 |
364 | /*output prior predictive samples for postprocessing for ABC*/
365 | fprintf(stdout,"\"Sample\",\"alpha\",\"delta\",\"mu\",\"G\",\"H\"\n");
366 | for (int j=0;j.
41 |
42 |
43 |
44 | ## Requirements
45 |
46 | 1. Intel Math Kernel Library (MKL) version >= 18 Update 1;
47 | 2. Intel C Compiler version >= 18.0.1 (Also compatible with GNU C Compiler version >= 7.0.0, but not tested);
48 | 3. CPU supporting Advance Vector Extensions instruction sets (either AVX, AVX2 or AVX512).
49 |
50 | ## Compile and run Benchmarks
51 |
52 | The build scripts and benchmarks assume a Linux Operating system using the Bourne-again shell
53 |
54 | 1. `cd path/to/example/`
55 | 2. `make`
56 | 3. `./run_bench.sh`
57 |
58 | Please note, the benchmark results presented in the Paper were performed using 4 cores of the Intel Xeon E5-2680v3 and 4 cores of the Intel Xeon Gold 6140. Speed-up factors may vary across CPU models, core counts and vector widths.
59 |
--------------------------------------------------------------------------------
/SMC_Inference_BEGE_model/Makefile:
--------------------------------------------------------------------------------
1 |
2 | all:
3 | make SMC_RW_LikeAnneal_BEGE_vec_par
4 | make SMC_RW_LikeAnneal_BEGE_novec_par
5 | make SMC_RW_LikeAnneal_BEGE_novec_nopar
6 | make SMC_RW_LikeAnneal_BEGE_vec_nopar
7 |
8 | SMC_RW_LikeAnneal_BEGE_vec_par: SMC_RW_LikeAnneal_BEGE_vec_par.c Makefile
9 | icc -mkl -qopenmp -O2 -xhost SMC_RW_LikeAnneal_BEGE_vec_par.c -o SMC_RW_LikeAnneal_BEGE_vec_par
10 |
11 | SMC_RW_LikeAnneal_BEGE_novec_par: SMC_RW_LikeAnneal_BEGE_novec_par.c Makefile
12 | icc -mkl -qopenmp -O2 -xhost -no-vec SMC_RW_LikeAnneal_BEGE_novec_par.c -o SMC_RW_LikeAnneal_BEGE_novec_par
13 |
14 | SMC_RW_LikeAnneal_BEGE_vec_nopar: SMC_RW_LikeAnneal_BEGE_vec_nopar.c Makefile
15 | icc -mkl -qopenmp -O2 -xhost SMC_RW_LikeAnneal_BEGE_vec_nopar.c -o SMC_RW_LikeAnneal_BEGE_vec_nopar
16 |
17 | SMC_RW_LikeAnneal_BEGE_novec_nopar: SMC_RW_LikeAnneal_BEGE_novec_nopar.c Makefile
18 | icc -mkl -O2 -xhost -no-vec SMC_RW_LikeAnneal_BEGE_novec_nopar.c -o SMC_RW_LikeAnneal_BEGE_novec_nopar
19 |
20 | clean:
21 | rm -f SMC_RW_LikeAnneal_BEGE_vec_par SMC_RW_LikeAnneal_BEGE_novec_par SMC_RW_LikeAnneal_BEGE_novec_nopar SMC_RW_LikeAnneal_BEGE_vec_nopar
22 |
--------------------------------------------------------------------------------
/SMC_Inference_BEGE_model/MonthlyReturns2018.csv:
--------------------------------------------------------------------------------
1 | 1099
2 | 0.0296
3 | 0.0264
4 | 0.0036
5 | -0.0324
6 | 0.0253
7 | 0.0262
8 | -0.0006
9 | 0.0418
10 | 0.0013
11 | 0.0046
12 | 0.0544
13 | -0.0234
14 | 0.0726
15 | 0.0197
16 | 0.0476
17 | -0.0431
18 | 0.0658
19 | 0.0209
20 | -0.0068
21 | -0.017
22 | 0.0881
23 | 0.0423
24 | 0.0152
25 | -0.0485
26 | 0.0062
27 | 0.0668
28 | 0.0288
29 | 0.0133
30 | 0.1181
31 | 0.0036
32 | 0.0466
33 | -0.0034
34 | -0.0089
35 | 0.0143
36 | -0.0639
37 | 0.097
38 | 0.0446
39 | 0.0818
40 | -0.0547
41 | -0.2012
42 | -0.1274
43 | 0.0133
44 | 0.0561
45 | 0.025
46 | 0.071
47 | -0.0206
48 | -0.0166
49 | -0.1627
50 | 0.0412
51 | 0.003
52 | -0.1275
53 | -0.0878
54 | -0.0304
55 | -0.0783
56 | 0.0624
57 | 0.1088
58 | -0.0643
59 | -0.0998
60 | -0.1324
61 | 0.139
62 | -0.0662
63 | 0.0041
64 | -0.2913
65 | 0.0804
66 | -0.0908
67 | -0.1353
68 | -0.0158
69 | 0.0546
70 | -0.1121
71 | -0.1796
72 | -0.2051
73 | -0.007
74 | 0.3384
75 | 0.3706
76 | -0.0294
77 | -0.1317
78 | -0.0588
79 | 0.044
80 | 0.0125
81 | -0.1524
82 | 0.0329
83 | 0.3885
84 | 0.2143
85 | 0.1311
86 | -0.0963
87 | 0.1205
88 | -0.1065
89 | -0.0836
90 | 0.0997
91 | 0.0183
92 | 0.126
93 | -0.025
94 | 0.0009
95 | -0.0179
96 | -0.0725
97 | 0.0264
98 | -0.1096
99 | 0.0558
100 | -0.0023
101 | -0.0166
102 | 0.0833
103 | 0.0036
104 | -0.0345
105 | -0.0194
106 | -0.0368
107 | 0.0906
108 | 0.0347
109 | 0.0593
110 | 0.0751
111 | 0.0265
112 | 0.0263
113 | 0.0703
114 | 0.0488
115 | 0.0456
116 | 0.0689
117 | 0.0249
118 | 0.0099
119 | -0.0814
120 | 0.0519
121 | 0.024
122 | 0.0667
123 | 0.0099
124 | 0.0098
125 | 0.0712
126 | 0.0327
127 | 0.0021
128 | 0.0335
129 | 0.0109
130 | -0.0027
131 | -0.0736
132 | -0.0083
133 | -0.0421
134 | 0.0891
135 | -0.0486
136 | -0.1361
137 | -0.0961
138 | -0.0831
139 | -0.0424
140 | 0.0049
141 | 0.0584
142 | -0.2382
143 | 0.1451
144 | -0.0383
145 | 0.2387
146 | 0.0734
147 | -0.0267
148 | 0.0081
149 | 0.078
150 | -0.0172
151 | 0.0419
152 | -0.0596
153 | 0.0351
154 | -0.1199
155 | -0.0018
156 | 0.068
157 | -0.0531
158 | 0.1024
159 | -0.0668
160 | 0.1688
161 | -0.0053
162 | -0.0362
163 | 0.0303
164 | -0.0241
165 | 0.0144
166 | 0.0205
167 | 0.0022
168 | -0.2195
169 | 0.0667
170 | 0.0316
171 | 0.0219
172 | 0.0239
173 | 0.0302
174 | -0.0161
175 | 0.0069
176 | -0.0417
177 | -0.0143
178 | 0.0084
179 | -0.0546
180 | 0.0139
181 | 0.0583
182 | 0.0587
183 | -0.0017
184 | -0.0087
185 | -0.0525
186 | -0.0192
187 | -0.0487
188 | 0.0079
189 | -0.0246
190 | -0.0658
191 | -0.0437
192 | 0.0594
193 | 0.0269
194 | 0.0351
195 | 0.018
196 | 0.0261
197 | 0.0682
198 | 0.0015
199 | 0.0512
200 | 0.0713
201 | 0.0615
202 | 0.0601
203 | 0.0081
204 | 0.0574
205 | 0.0182
206 | -0.0477
207 | 0.013
208 | 0.024
209 | -0.0115
210 | -0.0591
211 | 0.0636
212 | 0.0174
213 | 0.0037
214 | 0.0246
215 | -0.0169
216 | 0.0507
217 | 0.0549
218 | -0.0149
219 | 0.0157
220 | 0.0001
221 | 0.0016
222 | 0.0171
223 | 0.0403
224 | 0.0201
225 | 0.0623
226 | -0.0389
227 | 0.078
228 | 0.0173
229 | 0.0039
230 | -0.0217
231 | 0.062
232 | 0.0477
233 | 0.0389
234 | 0.0539
235 | 0.012
236 | 0.0624
237 | -0.0583
238 | 0.0587
239 | 0.0423
240 | 0.0393
241 | -0.0389
242 | -0.0269
243 | -0.0644
244 | -0.1017
245 | -0.0144
246 | -0.0001
247 | 0.0496
248 | 0.0125
249 | -0.0108
250 | -0.0167
251 | -0.048
252 | -0.0097
253 | 0.0529
254 | 0.0414
255 | -0.0174
256 | -0.0054
257 | 0.0247
258 | -0.0197
259 | 0.03
260 | -0.0393
261 | -0.0438
262 | 0.0807
263 | 0.0365
264 | 0.073
265 | -0.001
266 | -0.0509
267 | 0.0025
268 | -0.0297
269 | 0.0596
270 | -0.093
271 | 0.0326
272 | 0.0023
273 | -0.0293
274 | 0.0404
275 | -0.0187
276 | -0.0294
277 | 0.001
278 | 0.0554
279 | 0.026
280 | 0.0309
281 | 0.0314
282 | 0.0182
283 | 0.0513
284 | 0.017
285 | 0.0148
286 | 0.0126
287 | 0.0394
288 | 0.0431
289 | -0.0594
290 | 0.0136
291 | 0.0485
292 | 0.0481
293 | -0.0018
294 | 0.0276
295 | 0.0554
296 | 0.057
297 | 0.0141
298 | -0.0215
299 | 0.0486
300 | -0.0234
301 | -0.0262
302 | 0.0694
303 | 0.0427
304 | 0.007
305 | -0.0253
306 | 0.0057
307 | 0.0333
308 | 0.0145
309 | -0.0262
310 | 0.0444
311 | -0.0497
312 | 0.032
313 | 0.0383
314 | 0.0091
315 | -0.0076
316 | -0.0203
317 | -0.0066
318 | 0.0594
319 | 0.0293
320 | -0.0034
321 | -0.0027
322 | -0.0143
323 | -0.0283
324 | 0.0052
325 | -0.0189
326 | 0.024
327 | -0.0452
328 | 0.002
329 | 0.046
330 | 0.0283
331 | 0.0003
332 | 0.0513
333 | 0.0167
334 | 0.0365
335 | 0.0427
336 | 0.0309
337 | 0.0107
338 | 0.0499
339 | -0.0234
340 | 0.0639
341 | -0.0167
342 | 0.0938
343 | 0.0548
344 | 0.006
345 | 0.0302
346 | -0.0016
347 | 0.0311
348 | 0.0093
349 | 0.0655
350 | 0.019
351 | 0.0021
352 | -0.0036
353 | -0.0268
354 | 0.0703
355 | 0.0149
356 | -0.0303
357 | 0.0377
358 | 0.0664
359 | 0.0028
360 | -0.052
361 | 0.0348
362 | 0.0484
363 | -0.0318
364 | -0.0514
365 | 0.0052
366 | 0.0036
367 | 0.0316
368 | -0.0358
369 | -0.0206
370 | 0.0213
371 | 0.0426
372 | 0.0345
373 | -0.0074
374 | 0.0066
375 | -0.0511
376 | -0.0598
377 | -0.0432
378 | 0.023
379 | -0.0391
380 | 0.0466
381 | -0.0152
382 | 0.0327
383 | 0.0309
384 | 0.0231
385 | 0.0293
386 | 0.0439
387 | 0.0191
388 | 0.0466
389 | 0.0253
390 | 0.0301
391 | 0.0515
392 | 0.0071
393 | 0.0095
394 | 0.0028
395 | 0.0366
396 | 0.0173
397 | -0.0025
398 | 0.0317
399 | -0.0139
400 | -0.048
401 | 0.0128
402 | 0.016
403 | 0.0245
404 | -0.0698
405 | 0.0117
406 | -0.0163
407 | -0.0171
408 | 0.0312
409 | 0.0208
410 | -0.0237
411 | 0.0301
412 | -0.0599
413 | -0.0071
414 | 0.0469
415 | 0.0471
416 | 0.062
417 | 0.0357
418 | 0.0289
419 | 0.0029
420 | 0.024
421 | -0.0308
422 | 0.0283
423 | 0.0257
424 | -0.0215
425 | 0.0257
426 | 0.0445
427 | -0.0018
428 | -0.0387
429 | 0.0181
430 | -0.0068
431 | -0.0659
432 | -0.0865
433 | -0.0847
434 | 0.0628
435 | 0.0213
436 | -0.0522
437 | -0.0005
438 | 0.1087
439 | 0.0101
440 | 0.0493
441 | -0.0238
442 | 0.0308
443 | 0.0451
444 | 0.0176
445 | -0.02
446 | -0.0039
447 | 0.0507
448 | -0.0157
449 | 0.0253
450 | -0.0085
451 | 0.0183
452 | 0.0224
453 | 0.0154
454 | 0.0141
455 | 0.001
456 | 0.0142
457 | 0.0127
458 | 0.0174
459 | -0.0144
460 | 0.0269
461 | 0.0059
462 | 0
463 | 0.0003
464 | 0.0354
465 | 0.0044
466 | -0.0134
467 | 0.0311
468 | -0.0077
469 | -0.0551
470 | 0.0143
471 | 0.0273
472 | 0.0286
473 | 0.026
474 | -0.0003
475 | 0.0101
476 | 0.0072
477 | -0.0121
478 | -0.0251
479 | 0.0214
480 | -0.0566
481 | -0.0144
482 | -0.0163
483 | -0.0791
484 | -0.0106
485 | 0.0386
486 | 0.014
487 | 0.0013
488 | 0.0815
489 | 0.0078
490 | 0.0399
491 | 0.0389
492 | -0.0433
493 | 0.0241
494 | 0.0458
495 | -0.0089
496 | 0.0311
497 | -0.0309
498 | 0.0037
499 | 0.0305
500 | -0.0406
501 | -0.0375
502 | 0.002
503 | 0.0905
504 | 0.0228
505 | 0.0069
506 | -0.0272
507 | 0.0134
508 | 0.0403
509 | 0.0042
510 | 0.0543
511 | -0.0394
512 | -0.0125
513 | -0.0584
514 | 0.0264
515 | 0.0146
516 | -0.001
517 | -0.0718
518 | -0.07
519 | 0.0468
520 | -0.0298
521 | 0.0506
522 | -0.0379
523 | -0.0263
524 | -0.081
525 | 0.0513
526 | -0.0106
527 | -0.11
528 | -0.0692
529 | -0.0579
530 | 0.0693
531 | 0.0449
532 | 0.0418
533 | -0.0228
534 | 0.0459
535 | 0.0572
536 | 0.0484
537 | 0.0141
538 | 0.0413
539 | 0.0315
540 | -0.0398
541 | -0.001
542 | -0.045
543 | 0.0379
544 | -0.0085
545 | -0.0442
546 | -0.0046
547 | 0.0871
548 | 0.0249
549 | 0.0287
550 | 0.0063
551 | 0.0029
552 | 0.0125
553 | -0.0243
554 | -0.008
555 | 0.0326
556 | -0.0114
557 | 0.0052
558 | 0.046
559 | 0.0062
560 | -0.0329
561 | -0.0485
562 | -0.013
563 | -0.0568
564 | -0.0294
565 | -0.0156
566 | 0.0505
567 | -0.0382
568 | 0.0475
569 | -0.0083
570 | -0.1275
571 | 0.0061
572 | -0.0017
573 | -0.0047
574 | -0.0281
575 | -0.0529
576 | -0.0468
577 | -0.0283
578 | -0.0805
579 | -0.0935
580 | -0.1177
581 | 0.161
582 | -0.0451
583 | -0.0345
584 | 0.1366
585 | 0.0556
586 | 0.0266
587 | 0.0423
588 | 0.0519
589 | 0.0483
590 | -0.0659
591 | -0.0285
592 | -0.0426
593 | 0.0531
594 | 0.0265
595 | -0.016
596 | 0.1216
597 | 0.0032
598 | 0.0233
599 | -0.0149
600 | -0.0134
601 | 0.0405
602 | -0.0107
603 | -0.0056
604 | 0.0206
605 | -0.0242
606 | 0.0036
607 | 0.0565
608 | -0.0405
609 | -0.0195
610 | -0.0137
611 | 0.0015
612 | -0.0146
613 | 0.0471
614 | -0.0169
615 | -0.0175
616 | -0.0027
617 | -0.0438
618 | 0.04
619 | 0.0027
620 | -0.0601
621 | -0.0138
622 | 0.0285
623 | 0.0788
624 | 0.0176
625 | -0.0169
626 | 0.0511
627 | 0.0375
628 | -0.0143
629 | -0.1191
630 | 0.0271
631 | 0.0088
632 | 0.0423
633 | -0.0356
634 | 0.0568
635 | -0.0006
636 | -0.0221
637 | 0.0385
638 | 0.0082
639 | 0.0553
640 | -0.0082
641 | -0.081
642 | 0.0521
643 | 0.0179
644 | 0.0551
645 | -0.0122
646 | -0.129
647 | 0.0397
648 | 0.0526
649 | 0.0306
650 | 0.0649
651 | 0.018
652 | 0.0219
653 | 0.0106
654 | 0.0959
655 | -0.0452
656 | -0.0504
657 | 0.0056
658 | 0.0356
659 | -0.021
660 | 0.0011
661 | -0.0236
662 | -0.0154
663 | -0.0703
664 | -0.0717
665 | 0.0492
666 | 0.0336
667 | -0.0365
668 | -0.0324
669 | -0.0586
670 | -0.0187
671 | 0.0327
672 | -0.0399
673 | -0.0309
674 | -0.0319
675 | 0.1114
676 | 0.0129
677 | 0.113
678 | 0.0467
679 | 0.0055
680 | 0.036
681 | 0.0259
682 | 0.0282
683 | 0.0667
684 | 0.0052
685 | 0.0307
686 | -0.0407
687 | -0.005
688 | 0.0091
689 | -0.0344
690 | 0.0216
691 | -0.0178
692 | -0.0192
693 | -0.0482
694 | 0.0063
695 | -0.0052
696 | -0.0597
697 | 0.0182
698 | -0.0274
699 | 0.1028
700 | -0.008
701 | -0.0084
702 | -0.0176
703 | 0.0184
704 | 0.0799
705 | 0.0122
706 | -0.0084
707 | -0.0096
708 | 0.0509
709 | 0.0127
710 | -0.0074
711 | -0.0102
712 | -0.0454
713 | 0.0402
714 | 0.0648
715 | 0.0388
716 | 0.0065
717 | 0.0713
718 | 0.0488
719 | -0.0131
720 | 0.0462
721 | 0.0103
722 | -0.0645
723 | 0.0607
724 | -0.086
725 | 0.0466
726 | 0.0117
727 | -0.0327
728 | 0.1247
729 | 0.0439
730 | 0.0164
731 | -0.0211
732 | 0.0011
733 | 0.0394
734 | 0.0385
735 | 0.0352
736 | -0.0259
737 | -0.2324
738 | -0.0777
739 | 0.0681
740 | 0.0421
741 | 0.0475
742 | -0.0227
743 | 0.0056
744 | -0.0029
745 | 0.0479
746 | -0.0125
747 | -0.0331
748 | 0.033
749 | 0.0115
750 | -0.0229
751 | 0.0149
752 | 0.061
753 | -0.0225
754 | 0.0157
755 | 0.0433
756 | 0.0335
757 | -0.0135
758 | 0.072
759 | 0.0144
760 | -0.0076
761 | -0.0367
762 | 0.0103
763 | 0.0116
764 | -0.0785
765 | 0.0111
766 | 0.0183
767 | -0.0336
768 | 0.0842
769 | -0.0109
770 | -0.019
771 | -0.1015
772 | -0.0612
773 | -0.0192
774 | 0.0635
775 | 0.0246
776 | 0.0469
777 | 0.0719
778 | 0.0265
779 | -0.0028
780 | 0.0365
781 | -0.0494
782 | 0.0424
783 | 0.0232
784 | -0.0159
785 | 0.0129
786 | -0.0419
787 | 0.1084
788 | -0.0059
789 | 0.0109
790 | -0.0266
791 | 0.0107
792 | 0.003
793 | -0.0234
794 | 0.0377
795 | -0.0238
796 | 0.0119
797 | 0.0102
798 | 0.0413
799 | 0.0153
800 | 0.0093
801 | 0.0012
802 | 0.023
803 | -0.0305
804 | 0.0289
805 | 0.0031
806 | -0.0034
807 | 0.0371
808 | -0.0012
809 | 0.0141
810 | -0.0189
811 | 0.0165
812 | 0.0287
813 | -0.0255
814 | -0.0478
815 | 0.0068
816 | 0.0058
817 | -0.0303
818 | 0.0282
819 | 0.0401
820 | -0.0231
821 | 0.0134
822 | -0.0404
823 | 0.0086
824 | 0.018
825 | 0.0363
826 | 0.0219
827 | 0.0211
828 | 0.029
829 | 0.0272
830 | 0.0372
831 | 0.0055
832 | 0.0335
833 | -0.0152
834 | 0.0396
835 | 0.0103
836 | 0.0226
837 | 0.0133
838 | 0.0073
839 | 0.0206
840 | 0.0236
841 | -0.0114
842 | -0.0597
843 | 0.0277
844 | 0.0501
845 | 0.0086
846 | 0.0625
847 | -0.017
848 | 0.0498
849 | -0.0049
850 | -0.0502
851 | 0.0404
852 | 0.0674
853 | 0.041
854 | 0.0733
855 | -0.0415
856 | 0.0535
857 | -0.038
858 | 0.0298
859 | 0.0132
860 | 0.0015
861 | 0.0704
862 | 0.0476
863 | 0.0073
864 | -0.0307
865 | 0.0318
866 | -0.0246
867 | -0.1608
868 | 0.0615
869 | 0.0713
870 | 0.061
871 | 0.0616
872 | 0.035
873 | -0.0408
874 | 0.0345
875 | 0.0433
876 | -0.0246
877 | 0.0477
878 | -0.0347
879 | -0.0138
880 | -0.0281
881 | 0.0613
882 | 0.0337
883 | 0.0772
884 | -0.0474
885 | 0.0245
886 | 0.052
887 | -0.064
888 | -0.0442
889 | 0.0464
890 | -0.0251
891 | 0.0703
892 | -0.0545
893 | -0.0276
894 | -0.1072
895 | 0.0119
896 | 0.0313
897 | -0.1005
898 | -0.0726
899 | 0.0794
900 | 0.0072
901 | -0.0194
902 | -0.0213
903 | -0.0646
904 | -0.0925
905 | 0.0246
906 | 0.0754
907 | 0.0161
908 | -0.0144
909 | -0.0229
910 | 0.0424
911 | -0.052
912 | -0.0138
913 | -0.0721
914 | -0.0818
915 | 0.005
916 | -0.1035
917 | 0.0784
918 | 0.0596
919 | -0.0576
920 | -0.0257
921 | -0.0188
922 | 0.0109
923 | 0.0822
924 | 0.0605
925 | 0.0142
926 | 0.0235
927 | 0.0234
928 | -0.0124
929 | 0.0608
930 | 0.0135
931 | 0.0429
932 | 0.0215
933 | 0.014
934 | -0.0132
935 | -0.0183
936 | 0.0117
937 | 0.0186
938 | -0.0406
939 | 0.0008
940 | 0.016
941 | 0.0143
942 | 0.0454
943 | 0.0343
944 | -0.0276
945 | 0.0189
946 | -0.0197
947 | -0.0261
948 | 0.0365
949 | 0.0057
950 | 0.0392
951 | -0.0122
952 | 0.0049
953 | -0.0202
954 | 0.0361
955 | -0.0025
956 | 0.0304
957 | -0.003
958 | 0.0146
959 | 0.0073
960 | -0.0357
961 | -0.0035
962 | -0.0078
963 | 0.0203
964 | 0.0184
965 | 0.0323
966 | 0.0171
967 | 0.0087
968 | 0.014
969 | -0.0196
970 | 0.0068
971 | 0.0349
972 | 0.0324
973 | -0.0196
974 | -0.0373
975 | 0.0092
976 | 0.0322
977 | 0.018
978 | -0.0483
979 | -0.0087
980 | -0.0636
981 | -0.0309
982 | -0.0093
983 | 0.046
984 | 0.0186
985 | -0.0844
986 | -0.0077
987 | 0.0153
988 | -0.0924
989 | -0.1723
990 | -0.0786
991 | 0.0174
992 | -0.0812
993 | -0.101
994 | 0.0895
995 | 0.1019
996 | 0.0521
997 | 0.0043
998 | 0.0772
999 | 0.0333
1000 | 0.0408
1001 | -0.0259
1002 | 0.0556
1003 | 0.0275
1004 | -0.0336
1005 | 0.034
1006 | 0.0631
1007 | 0.02
1008 | -0.0789
1009 | -0.0556
1010 | 0.0693
1011 | -0.0477
1012 | 0.0954
1013 | 0.0388
1014 | 0.006
1015 | 0.0682
1016 | 0.0199
1017 | 0.0349
1018 | 0.0045
1019 | 0.029
1020 | -0.0127
1021 | -0.0175
1022 | -0.0236
1023 | -0.0599
1024 | -0.0759
1025 | 0.1135
1026 | -0.0028
1027 | 0.0074
1028 | 0.0505
1029 | 0.0442
1030 | 0.0311
1031 | -0.0085
1032 | -0.0619
1033 | 0.0389
1034 | 0.0079
1035 | 0.0255
1036 | 0.0273
1037 | -0.0176
1038 | 0.0078
1039 | 0.0118
1040 | 0.0557
1041 | 0.0129
1042 | 0.0403
1043 | 0.0155
1044 | 0.028
1045 | -0.012
1046 | 0.0565
1047 | -0.0271
1048 | 0.0377
1049 | 0.0418
1050 | 0.0312
1051 | 0.0281
1052 | -0.0332
1053 | 0.0465
1054 | 0.0043
1055 | -0.0019
1056 | 0.0206
1057 | 0.0261
1058 | -0.0204
1059 | 0.0424
1060 | -0.0197
1061 | 0.0252
1062 | 0.0255
1063 | -0.0006
1064 | -0.0311
1065 | 0.0613
1066 | -0.0112
1067 | 0.0059
1068 | 0.0136
1069 | -0.0153
1070 | 0.0154
1071 | -0.0604
1072 | -0.0308
1073 | 0.0775
1074 | 0.0056
1075 | -0.0217
1076 | -0.0577
1077 | -0.0007
1078 | 0.0696
1079 | 0.0092
1080 | 0.0178
1081 | -0.0005
1082 | 0.0395
1083 | 0.005
1084 | 0.0025
1085 | -0.0202
1086 | 0.0486
1087 | 0.0182
1088 | 0.0194
1089 | 0.0357
1090 | 0.0017
1091 | 0.0109
1092 | 0.0106
1093 | 0.0078
1094 | 0.0187
1095 | 0.0016
1096 | 0.0251
1097 | 0.0225
1098 | 0.0312
1099 | 0.0106
1100 | 0.0557
1101 |
--------------------------------------------------------------------------------
/SMC_Inference_BEGE_model/SMC_RW_LikeAnneal_BEGE_novec_nopar.c:
--------------------------------------------------------------------------------
1 | /**
2 | * @file SMC_RW_LikeAnneal_BEGE.c
3 | * @brief Demonstration of vectorisationn and parallelisation for inference on
4 | * the Bad Environment, Good Evironment (BEGE) model.
5 | *
6 | * @details Random walk adaptive SMC with likelihood annealing with an expensive
7 | * likelihood.
8 | *
9 | * @author Chris C. Drovandi (c.drovandi@qut.edu.au)
10 | * School of Mathematical Sciences
11 | * Queensland University of Technology
12 | * @author David J. Warne (david.warne@qut.edu.au)
13 | * School of Mathematical Sciences
14 | * Queensland University of Technology
15 | *
16 | * @date 4 Feb 2019
17 | */
18 |
19 | #include
20 | #include
21 | #include
22 |
23 | /* Intel headers */
24 | #include
25 | #include
26 | #include
27 |
28 | /* OpenMP header */
29 | #include
30 |
31 | /* length of vector processing units and ideal memory alignement*/
32 | #if defined(__AVX512BW__)
33 | #define VECLEN 8
34 | #define ALIGN 64
35 | #elif defined(__AVX2__)
36 | #define VECLEN 4
37 | #define ALIGN 64
38 | #elif defined(__AVX__)
39 | #define VECLEN 4
40 | #define ALIGN 32
41 | #elif defined(__SSE4_2__)
42 | #define VECLEN 2
43 | #define ALIGN 16
44 | #endif
45 |
46 |
47 | #define ITMAX 100 /*maximum allowed number of iterations*/
48 | #define EPS 3.0e-7 /*relative accuracy*/
49 |
50 | #define FPMIN 1.0e-30 /*number near the smallest representable
51 | floating point number*/
52 |
53 | #define NUM_PARAM 11
54 |
55 | /*function prototypes for gamma function and incomplete gamma functions*/
56 |
57 | /* error message function*/
58 | void
59 | nrerror(char *);
60 |
61 | /* series approximation @note could be a challenge to vectorise */
62 | void
63 | gser_vec(double * restrict, double * restrict, double * restrict, double * restrict);
64 |
65 | /* compute P(a,x) = \gamma(a,x)/Gamma(a)*/
66 | void
67 | gammp_vec(double, double, int, double * restrict, double * restrict);
68 |
69 | /* compute Q(a,x) = 1 - P(a,x) = \Gamma(a,x)/Gamma(a)*/
70 | void
71 | gammq_vec(double, double, int, double * restrict, double * restrict);
72 |
73 | /* compute ln Gamma(a) @note possibly use the intel imf lgamma*/
74 | void
75 | gammln_vec(double * restrict, double * restrict);
76 |
77 | double*
78 | loadData(char *, int *);
79 |
80 | double
81 | compute_ESS_diff(double, double, double * restrict, double * restrict,
82 | unsigned int);
83 |
84 | double
85 | SMC_RW_LikeAnneal(VSLStreamStatePtr, unsigned int, int,double* restrict,
86 | double * restrict, double * restrict, double* restrict);
87 |
88 | double
89 | log_prior(double * restrict, double * restrict, double * restrict);
90 |
91 | double
92 | bege_gjrgarch_likelihood(double * restrict, int, double * restrict,
93 | double * restrict, double * restrict);
94 | double
95 | loglikedgam_vec(double, double, double, double, double, double);
96 |
97 | double
98 | quantile(unsigned int, double *,double);
99 |
100 | /**
101 | * @brief main entry point of program
102 | */
103 | int
104 | main(int argc, char ** argv)
105 | {
106 | char *filename;
107 | double *rate_return, *theta;
108 | int ndat;
109 | unsigned int seed;
110 | int N, NUM_PARTICLES;
111 |
112 | /* Intel MKL VSL random stream */
113 | VSLStreamStatePtr stream;
114 |
115 | /*limits for uniform priors*/
116 | __declspec(align(ALIGN)) double prior_l[NUM_PARAM] = {1e-4,1e-4,1e-4,1e-4,
117 | 1e-4,1e-4,1e-4,1e-4,-0.2,1e-4,-0.9};
118 | __declspec(align(ALIGN)) double prior_u[NUM_PARAM] = { 0.5, 0.3,0.99, 0.5,
119 | 0.5, 1.0, 0.3,0.99, 0.1,0.75, 0.9};
120 |
121 | if (argc < 4)
122 | {
123 | fprintf(stderr,"Usage: [%s] N datafile seed\n",argv[0]);
124 | exit(1);
125 | }
126 | NUM_PARTICLES = (int)atoi(argv[1]);
127 | filename = argv[2];
128 | seed = (unsigned int)atoi(argv[3]);
129 |
130 | /* load finance data */
131 | rate_return = loadData(filename, &ndat);
132 |
133 | /* to ensure that particles pack neatly into vectors with no leftovers*/
134 | if (NUM_PARTICLES%VECLEN == 0)
135 | {
136 | N = NUM_PARTICLES;
137 | }
138 | else
139 | {
140 | N = ((NUM_PARTICLES / VECLEN) + 1)*VECLEN;
141 | }
142 | fprintf(stderr,"Particles: %d, Vector length: %d\n",N,VECLEN);
143 |
144 | /* allocate memory for particles */ /**@note theta[N][NUM_PARAM]*/
145 | theta = (double*)_mm_malloc(NUM_PARAM*N*sizeof(double),ALIGN);
146 |
147 | /*initialise RNG */
148 | vslNewStream(&stream,VSL_BRNG_MT2203,seed);
149 |
150 | /* perform Random Walk SMC with Likelihood Annealling*/
151 | SMC_RW_LikeAnneal(stream, N, ndat, rate_return, prior_l, prior_u, theta);
152 |
153 | /*output particles*/
154 | for (int i=0;i w_max) ? logw[i] : w_max;
481 | }
482 | w_sum = 0;
483 | w_sum2 = 0;
484 | for (int i=0;i= Nf/2.0)
500 | {
501 | gamma_tp1 = 1.0;
502 | }
503 | else
504 | {
505 | double a = 0;
506 | double b = 0;
507 | double p = 0;
508 | double fa = 0;
509 | double fp = 0;
510 | double err = 0;
511 |
512 | /* find optimum temperature stepp using bisection method */
513 | a = gamma_t + 1e-6;
514 | b = 1.0;
515 | p = (a + b)/2.0;
516 | fp = compute_ESS_diff(p,gamma_t,w_temp,loglike,N);
517 | err = fabs(fp);
518 | while (err > 1e-5)
519 | {
520 | fa = compute_ESS_diff(a,gamma_t,w_temp,loglike,N);
521 | if (fa*fp < 0)
522 | {
523 | b = p;
524 | }
525 | else
526 | {
527 | a = p;
528 | }
529 | p = (a + b)/2.0;
530 | fp = compute_ESS_diff(p,gamma_t,w_temp,loglike,N);
531 | err = fabs(fp);
532 | }
533 | gamma_tp1 = p;
534 | }
535 |
536 | fprintf(stderr,"gamma(t) = %g next gamma(t+1) = %g\n",gamma_t,gamma_tp1);
537 | /* substitute the value of just calculated gamma */
538 | for (int i=0;i w_max) ? logw[i] : w_max;
547 | }
548 | w_sum = 0;
549 | w_sum2 = 0;
550 | for (int i=0;i w_sum3 && k < N-1)
573 | {
574 | k++;
575 | w_sum3 += w[k];
576 | }
577 | r[i] = k;
578 | }
579 |
580 | /* re-assign particles*/
581 | for (int i=0;i median_ESJD[ind])
740 | {
741 | ind = j;
742 | }
743 | }
744 |
745 | h_opt = h[ind];
746 | fprintf(stderr,"The scale is %f\n",h_opt);
747 |
748 | memset(dist_move,0,N*sizeof(double));
749 | belowThreshold = 1;
750 | R_move = 0;
751 |
752 |
753 | fprintf(stderr,"MCMC proposals for particle mutation...\n");
754 | /* Performing remaining repeats */
755 | while (belowThreshold)
756 | {
757 | int sum_cond = 0;;
758 | R_move++;
759 | /*generate N MCMC proposals*/
760 | memset(mu,0,NUM_PARAM*sizeof(double));
761 | vdRngGaussianMV(VSL_RNG_METHOD_GAUSSIANMV_BOXMULLER2,stream,N,
762 | theta_particle_prop, NUM_PARAM, VSL_MATRIX_STORAGE_FULL,mu,T);
763 | /* N uniforms for accept/reject step*/
764 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,N,u,0,1);
765 | for (int i=0;i median_dist);
820 | }
821 | belowThreshold = (sum_cond < (int)ceil(Nf*0.5));
822 | }
823 | fprintf(stderr,"The value of R_move was %d\n",R_move);
824 |
825 | gamma_t = gamma_tp1;
826 | }
827 |
828 | /* transform theta*/
829 | for (int i=0;i 0) ? sum[i]*exp(-x[i] + a[i]*log(x[i]) - (gln[i])) : 0.0;
1156 | }
1157 | return;
1158 | }
1159 | }
1160 |
1161 |
1162 | /**
1163 | * @brief computes P(a,x[1:N]/b) = \gamma(a,x[1:N]/b)/\Gamma(a)
1164 | * @note assumes the array x contains a monotonically increasing sequence.
1165 | */
1166 | void
1167 | gammp_vec(double a, double b, int N, double * restrict x, double * restrict P)
1168 | {
1169 | __declspec(align(ALIGN)) double _a[VECLEN];
1170 | for(int i=0;i= 0 && x[j] > A)
1239 | {
1240 | x[j+1] = x[j];
1241 | j--;
1242 | }
1243 | x[j+1] = A;
1244 | }
1245 | }
1246 |
1247 | /**
1248 | * @brief compute the q-quantile
1249 | *
1250 | * @param n number of samples
1251 | * @param x samples
1252 | * @param q the q-th quantile level
1253 | * @return the value of quantile
1254 | */
1255 | double
1256 | quantile(unsigned int n, double *x,double q)
1257 | {
1258 | double u,l;
1259 | unsigned int i;
1260 | /*sort samples */
1261 | insertionSort(n,x);
1262 |
1263 | /* pick the samples closes to the q quantile*/
1264 |
1265 | /* upper and lower bound of the quantile value*/
1266 | u = ceil(((double)n)*q);
1267 | l = floor(((double)n)*q);
1268 |
1269 | /* nearest neighbour interpolation */
1270 | i = (unsigned int)((u - ((double)n)*q < ((double)n)*q - l) ? u : l);
1271 | return x[i];
1272 | }
1273 |
1274 |
1275 |
--------------------------------------------------------------------------------
/SMC_Inference_BEGE_model/SMC_RW_LikeAnneal_BEGE_vec_nopar.c:
--------------------------------------------------------------------------------
1 | /**
2 | * @file SMC_RW_LikeAnneal_BEGE.c
3 | * @brief Demonstration of vectorisationn and parallelisation for inference on
4 | * the Bad Environment, Good Evironment (BEGE) model.
5 | *
6 | * @details Random walk adaptive SMC with likelihood annealing with an expensive
7 | * likelihood.
8 | *
9 | * @author Chris C. Drovandi (c.drovandi@qut.edu.au)
10 | * School of Mathematical Sciences
11 | * Queensland University of Technology
12 | * @author David J. Warne (david.warne@qut.edu.au)
13 | * School of Mathematical Sciences
14 | * Queensland University of Technology
15 | *
16 | * @date 4 Feb 2019
17 | */
18 |
19 | #include
20 | #include
21 | #include
22 |
23 | /* Intel headers */
24 | #include
25 | #include
26 | #include
27 |
28 | /* OpenMP header */
29 | #include
30 |
31 | /* length of vector processing units and ideal memory alignement*/
32 | #if defined(__AVX512BW__)
33 | #define VECLEN 8
34 | #define ALIGN 64
35 | #elif defined(__AVX2__)
36 | #define VECLEN 4
37 | #define ALIGN 64
38 | #elif defined(__AVX__)
39 | #define VECLEN 4
40 | #define ALIGN 32
41 | #elif defined(__SSE4_2__)
42 | #define VECLEN 2
43 | #define ALIGN 16
44 | #endif
45 |
46 |
47 | #define ITMAX 100 /*maximum allowed number of iterations*/
48 | #define EPS 3.0e-7 /*relative accuracy*/
49 |
50 | #define FPMIN 1.0e-30 /*number near the smallest representable
51 | floating point number*/
52 |
53 | #define NUM_PARAM 11
54 |
55 | /*function prototypes for gamma function and incomplete gamma functions*/
56 |
57 | /* error message function*/
58 | void
59 | nrerror(char *);
60 |
61 | /* series approximation @note could be a challenge to vectorise */
62 | void
63 | gser_vec(double * restrict, double * restrict, double * restrict, double * restrict);
64 |
65 | /* compute P(a,x) = \gamma(a,x)/Gamma(a)*/
66 | void
67 | gammp_vec(double, double, int, double * restrict, double * restrict);
68 |
69 | /* compute Q(a,x) = 1 - P(a,x) = \Gamma(a,x)/Gamma(a)*/
70 | void
71 | gammq_vec(double, double, int, double * restrict, double * restrict);
72 |
73 | /* compute ln Gamma(a) @note possibly use the intel imf lgamma*/
74 | void
75 | gammln_vec(double * restrict, double * restrict);
76 |
77 | double*
78 | loadData(char *, int *);
79 |
80 | double
81 | compute_ESS_diff(double, double, double * restrict, double * restrict,
82 | unsigned int);
83 |
84 | double
85 | SMC_RW_LikeAnneal(VSLStreamStatePtr, unsigned int, int,double* restrict,
86 | double * restrict, double * restrict, double* restrict);
87 |
88 | double
89 | log_prior(double * restrict, double * restrict, double * restrict);
90 |
91 | double
92 | bege_gjrgarch_likelihood(double * restrict, int, double * restrict,
93 | double * restrict, double * restrict);
94 | double
95 | loglikedgam_vec(double, double, double, double, double, double);
96 |
97 | double
98 | quantile(unsigned int, double *,double);
99 |
100 | /**
101 | * @brief main entry point of program
102 | */
103 | int
104 | main(int argc, char ** argv)
105 | {
106 | char *filename;
107 | double *rate_return, *theta;
108 | int ndat;
109 | unsigned int seed;
110 | int N, NUM_PARTICLES;
111 |
112 | /* Intel MKL VSL random stream */
113 | VSLStreamStatePtr stream;
114 |
115 | /*limits for uniform priors*/
116 | __declspec(align(ALIGN)) double prior_l[NUM_PARAM] = {1e-4,1e-4,1e-4,1e-4,
117 | 1e-4,1e-4,1e-4,1e-4,-0.2,1e-4,-0.9};
118 | __declspec(align(ALIGN)) double prior_u[NUM_PARAM] = { 0.5, 0.3,0.99, 0.5,
119 | 0.5, 1.0, 0.3,0.99, 0.1,0.75, 0.9};
120 |
121 | if (argc < 4)
122 | {
123 | fprintf(stderr,"Usage: [%s] N datafile seed\n",argv[0]);
124 | exit(1);
125 | }
126 | NUM_PARTICLES = (int)atoi(argv[1]);
127 | filename = argv[2];
128 | seed = (unsigned int)atoi(argv[3]);
129 |
130 | /* load finance data */
131 | rate_return = loadData(filename, &ndat);
132 |
133 | /* to ensure that particles pack neatly into vectors with no leftovers*/
134 | if (NUM_PARTICLES%VECLEN == 0)
135 | {
136 | N = NUM_PARTICLES;
137 | }
138 | else
139 | {
140 | N = ((NUM_PARTICLES / VECLEN) + 1)*VECLEN;
141 | }
142 | fprintf(stderr,"Particles: %d, Vector length: %d\n",N,VECLEN);
143 |
144 | /* allocate memory for particles */ /**@note theta[N][NUM_PARAM]*/
145 | theta = (double*)_mm_malloc(NUM_PARAM*N*sizeof(double),ALIGN);
146 |
147 | /*initialise RNG */
148 | vslNewStream(&stream,VSL_BRNG_MT2203,seed);
149 |
150 | /* perform Random Walk SMC with Likelihood Annealling*/
151 | SMC_RW_LikeAnneal(stream, N, ndat, rate_return, prior_l, prior_u, theta);
152 |
153 | /*output particles*/
154 | for (int i=0;i w_max) ? logw[i] : w_max;
487 | }
488 | w_sum = 0;
489 | w_sum2 = 0;
490 | #pragma omp simd reduction (+:w_sum)
491 | for (int i=0;i= Nf/2.0)
508 | {
509 | gamma_tp1 = 1.0;
510 | }
511 | else
512 | {
513 | double a = 0;
514 | double b = 0;
515 | double p = 0;
516 | double fa = 0;
517 | double fp = 0;
518 | double err = 0;
519 |
520 | /* find optimum temperature stepp using bisection method */
521 | a = gamma_t + 1e-6;
522 | b = 1.0;
523 | p = (a + b)/2.0;
524 | fp = compute_ESS_diff(p,gamma_t,w_temp,loglike,N);
525 | err = fabs(fp);
526 | while (err > 1e-5)
527 | {
528 | fa = compute_ESS_diff(a,gamma_t,w_temp,loglike,N);
529 | if (fa*fp < 0)
530 | {
531 | b = p;
532 | }
533 | else
534 | {
535 | a = p;
536 | }
537 | p = (a + b)/2.0;
538 | fp = compute_ESS_diff(p,gamma_t,w_temp,loglike,N);
539 | err = fabs(fp);
540 | }
541 | gamma_tp1 = p;
542 | }
543 |
544 | fprintf(stderr,"gamma(t) = %g next gamma(t+1) = %g\n",gamma_t,gamma_tp1);
545 | /* substitute the value of just calculated gamma */
546 | #pragma omp simd
547 | for (int i=0;i w_max) ? logw[i] : w_max;
556 | }
557 | w_sum = 0;
558 | w_sum2 = 0;
559 | #pragma omp simd reduction (+:w_sum)
560 | for (int i=0;i w_sum3 && k < N-1)
583 | {
584 | k++;
585 | w_sum3 += w[k];
586 | }
587 | r[i] = k;
588 | }
589 |
590 | /* re-assign particles*/
591 | for (int i=0;i median_ESJD[ind])
754 | {
755 | ind = j;
756 | }
757 | }
758 |
759 | h_opt = h[ind];
760 | fprintf(stderr,"The scale is %f\n",h_opt);
761 |
762 | memset(dist_move,0,N*sizeof(double));
763 | belowThreshold = 1;
764 | R_move = 0;
765 |
766 |
767 | fprintf(stderr,"MCMC proposals for particle mutation...\n");
768 | /* Performing remaining repeats */
769 | while (belowThreshold)
770 | {
771 | int sum_cond = 0;;
772 | R_move++;
773 | /*generate N MCMC proposals*/
774 | memset(mu,0,NUM_PARAM*sizeof(double));
775 | vdRngGaussianMV(VSL_RNG_METHOD_GAUSSIANMV_BOXMULLER2,stream,N,
776 | theta_particle_prop, NUM_PARAM, VSL_MATRIX_STORAGE_FULL,mu,T);
777 | /* N uniforms for accept/reject step*/
778 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,N,u,0,1);
779 | for (int i=0;i median_dist);
837 | }
838 | belowThreshold = (sum_cond < (int)ceil(Nf*0.5));
839 | }
840 | fprintf(stderr,"The value of R_move was %d\n",R_move);
841 |
842 | gamma_t = gamma_tp1;
843 | }
844 |
845 | /* transform theta*/
846 | for (int i=0;i 0) ? sum[i]*exp(-x[i] + a[i]*log(x[i]) - (gln[i])) : 0.0;
1192 | }
1193 | return;
1194 | }
1195 | }
1196 |
1197 |
1198 | /**
1199 | * @brief computes P(a,x[1:N]/b) = \gamma(a,x[1:N]/b)/\Gamma(a)
1200 | * @note assumes the array x contains a monotonically increasing sequence.
1201 | */
1202 | void
1203 | gammp_vec(double a, double b, int N, double * restrict x, double * restrict P)
1204 | {
1205 | __declspec(align(ALIGN)) double _a[VECLEN];
1206 | #pragma omp simd
1207 | for(int i=0;i= 0 && x[j] > A)
1281 | {
1282 | x[j+1] = x[j];
1283 | j--;
1284 | }
1285 | x[j+1] = A;
1286 | }
1287 | }
1288 |
1289 | /**
1290 | * @brief compute the q-quantile
1291 | *
1292 | * @param n number of samples
1293 | * @param x samples
1294 | * @param q the q-th quantile level
1295 | * @return the value of quantile
1296 | */
1297 | double
1298 | quantile(unsigned int n, double *x,double q)
1299 | {
1300 | double u,l;
1301 | unsigned int i;
1302 | /*sort samples */
1303 | insertionSort(n,x);
1304 |
1305 | /* pick the samples closes to the q quantile*/
1306 |
1307 | /* upper and lower bound of the quantile value*/
1308 | u = ceil(((double)n)*q);
1309 | l = floor(((double)n)*q);
1310 |
1311 | /* nearest neighbour interpolation */
1312 | i = (unsigned int)((u - ((double)n)*q < ((double)n)*q - l) ? u : l);
1313 | return x[i];
1314 | }
1315 |
1316 |
1317 |
--------------------------------------------------------------------------------
/SMC_Inference_BEGE_model/run_bench.sh:
--------------------------------------------------------------------------------
1 | #!/bin/bash
2 |
3 | DAT=MonthlyReturns2018.csv
4 | SEED=1337
5 |
6 | echo "Prior weak informativity test benchmark"
7 |
8 | #run and time vectorised parallel version
9 | echo 'timing parallel+SIMD'
10 | time ./weak_info_test_vec_par $N $K $SEED > /dev/null
11 | echo 'done'
12 |
13 | #run and time scalar parallel version
14 | # (still uses OpenMP and MKL/VSL, but SIMD and auto-vectorisation disabled)
15 | echo 'timing parallel+scalar'
16 | time ./weak_info_test_novec_par $N $K $SEED > /dev/null
17 | echo 'done'
18 |
19 | #run and time scalar sequential version
20 | # (still uses MKL/VSL, but OpenMP and auto-vectorisation disabled )
21 | echo 'timing sequential+scalar'
22 | time ./weak_info_test_novec_nopar $N $K $SEED > /dev/null
23 | echo 'done'
24 |
--------------------------------------------------------------------------------
/Weakly_Informative_Priors/Makefile:
--------------------------------------------------------------------------------
1 |
2 | all:
3 | make weak_info_test_vec_par
4 | make weak_info_test_novec_nopar
5 | make weak_info_test_novec_par
6 |
7 | weak_info_test_vec_par: weak_info_test_vec_par.c Makefile
8 | icc -mkl -qopenmp -O2 -xhost weak_info_test_vec_par.c -o weak_info_test_vec_par
9 |
10 | weak_info_test_novec_par: weak_info_test_vec_par.c Makefile
11 | icc -mkl -qopenmp -O2 -xhost -qno-openmp-simd -no-vec weak_info_test_vec_par.c -o weak_info_test_novec_par
12 |
13 | weak_info_test_novec_nopar: weak_info_test_vec_par.c Makefile
14 | icc -mkl -O2 -xhost -no-vec weak_info_test_vec_par.c -o weak_info_test_novec_nopar
15 |
16 | clean:
17 | rm -f weak_info_test_vec_par weak_info_test_novec_nopar weak_info_test_novec_par
18 |
--------------------------------------------------------------------------------
/Weakly_Informative_Priors/run_bench.sh:
--------------------------------------------------------------------------------
1 | #!/bin/bash
2 |
3 | N=400
4 | K=400
5 | SEED=1337
6 |
7 | echo "Prior weak informativity test benchmark"
8 | echo "N = $N K = $K"
9 |
10 | #run and time vectorised parallel version
11 | echo 'timing parallel+SIMD'
12 | time ./weak_info_test_vec_par $N $K $SEED > /dev/null
13 | echo 'done'
14 |
15 | #run and time scalar parallel version
16 | # (still uses OpenMP and MKL/VSL, but SIMD and auto-vectorisation disabled)
17 | echo 'timing parallel+scalar'
18 | time ./weak_info_test_novec_par $N $K $SEED > /dev/null
19 | echo 'done'
20 |
21 | #run and time scalar sequential version
22 | # (still uses MKL/VSL, but OpenMP and auto-vectorisation disabled )
23 | echo 'timing sequential+scalar'
24 | time ./weak_info_test_novec_nopar $N $K $SEED > /dev/null
25 | echo 'done'
26 |
--------------------------------------------------------------------------------
/Weakly_Informative_Priors/weak_info_test_vec_par.c:
--------------------------------------------------------------------------------
1 | /* Bayesian computations using SIMD operations
2 | * Copyright (C) 2019 David J. Warne, Christopher C. Drovandi, Scott A. Sisson
3 | *
4 | * This program is free software: you can redistribute it and/or modify
5 | * it under the terms of the GNU General Public License as published by
6 | * the Free Software Foundation, either version 3 of the License, or
7 | * (at your option) any later version.
8 | *
9 | * This program is distributed in the hope that it will be useful,
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 | * GNU General Public License for more details.
13 | *
14 | * You should have received a copy of the GNU General Public License
15 | * along with this program. If not, see .
16 | */
17 | /**
18 | * @file weakInfo_bioassy.c
19 | * @brief Demonstration of Intel Xeon Phi weakly informative prior selection.
20 | * @details The approach is based on the prior predictive p-value aproach
21 | * applied to the bioassy application.
22 | *
23 | * @author Chris C. Drovandi (c.drovandi@qut.edu.au)
24 | * School of Mathematical Sciences
25 | * Queensland University of Technology
26 | *
27 | * @author David J. Warne (david.warne@qut.edu.au)
28 | * School of Mathematical Sciences
29 | * Queensland University of Technology
30 | *
31 | * @date 13 Nov 2018
32 | */
33 |
34 | /* standard C headers */
35 | #include
36 | #include
37 | #include
38 | #include
39 |
40 | /* Intel headers */
41 | #include
42 | #include
43 |
44 | /* OpenMP header */
45 | #include
46 |
47 | /* length of vector processing units and ideal memory alignement*/
48 | #if defined(__AVX512BW__)
49 | #define VECLEN 8
50 | #define ALIGN 64
51 | #elif defined(__AVX2__)
52 | #define VECLEN 4
53 | #define ALIGN 64
54 | #elif defined(__AVX__)
55 | #define VECLEN 4
56 | #define ALIGN 32
57 | #elif defined(__SSE4_2__)
58 | #define VECLEN 2
59 | #define ALIGN 16
60 | #endif
61 |
62 | /* macro definitions*/
63 | #define SIZE_D 4
64 | #define SIGMA0_BASE 10.0
65 | #define SIGMA1_BASE 2.5
66 | #define NUM_PARTICLES 500
67 |
68 | /* function prototype declarations*/
69 | void
70 | compute_pvalues(VSLStreamStatePtr, double, double, unsigned int,
71 | double * restrict, double * restrict);
72 | void
73 | simulate_bioassay(VSLStreamStatePtr, int * restrict, double , double,
74 | double * restrict);
75 | unsigned int
76 | nCk(unsigned int, unsigned int);
77 | void
78 | loglike_bioassay(unsigned int, double * restrict, int * restrict,
79 | double * restrict, double * restrict);
80 | void
81 | bvnpdf(unsigned int, double * restrict, double * restrict, double * restrict,
82 | double * restrict);
83 | double
84 | SMC_RW(VSLStreamStatePtr, unsigned int, int* restrict, double* restrict, double,
85 | double);
86 | double
87 | quantile(unsigned int, double * restrict, double);
88 | void
89 | insertionSort(unsigned int, double * restrict);
90 | double
91 | compute_ESS_diff(double, double, double * restrict, double * restrict,
92 | unsigned int);
93 | double
94 | logsumexp(double * restrict, unsigned int );
95 |
96 |
97 | /**
98 | * @brief program entry point
99 | *
100 | * @details computes weak information test for a set of prior hyperparameters.
101 | * p-values computations are distributed across cores. Each P-value calculation
102 | * further parallelised through SIMD operations withing the SMC step.
103 | *
104 | * @param argc number of command line arguments
105 | * @param argv vector of argument strings
106 | */
107 | int
108 | main(int argc, char ** argv)
109 | {
110 | /*dose levels (standardised)*/
111 | __declspec(align(ALIGN)) double d[SIZE_D];
112 | unsigned int num_datasets, sims, seed;
113 |
114 | /*prior predictive p-values*/
115 | double *pvals;
116 |
117 | /*our hyperparameters*/
118 | double *sigma0, *sigma1;
119 |
120 | /*Intel MKL VSL random stream */
121 | VSLStreamStatePtr stream;
122 |
123 | /*initialise RNG stream for sampling hyperparameter space*/
124 | vslNewStream(&stream,VSL_BRNG_MT19937,1337);
125 |
126 | /* dose data*/
127 | d[0] = -0.86;
128 | d[1] = -0.30;
129 | d[2] = -0.05;
130 | d[3] = -0.73;
131 |
132 | /* read commandline arguments*/
133 | if (argc < 4)
134 | {
135 | fprintf(stderr,"Usage : [%s] numdatasets sims seed\n",argv[0]);
136 | exit(1);
137 | }
138 | num_datasets = (unsigned int)atoi(argv[1]);
139 | sims = (unsigned int)atoi(argv[2]);
140 | seed = (unsigned int)atoi(argv[3]);
141 |
142 | /* allcote aligned memory for hyperparameters and p-values*/
143 | sigma0 = (double *)_mm_malloc(sims*sizeof(double),ALIGN);
144 | sigma1 = (double *)_mm_malloc(sims*sizeof(double),ALIGN);
145 | pvals = (double *)_mm_malloc(sims*sizeof(double),ALIGN);
146 |
147 | /* sample hyperparameter space*/
148 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,sims,sigma0,0.1,10.0);
149 | vdRngUniform(VSL_RNG_METHOD_UNIFORM_STD,stream,sims,sigma1,0.1,20.0);
150 |
151 | /* the base prior representing current best information is appended*/
152 | sigma0[sims-1] = SIGMA0_BASE;
153 | sigma1[sims-1] = SIGMA1_BASE;
154 |
155 | /* create threads */
156 | #pragma omp parallel shared(num_datasets,sims,sigma0,sigma1,pvals,d)
157 | {
158 | /*Intel MKL VSL random stream*/
159 | VSLStreamStatePtr stream_sims;
160 | int thread_id, num_threads;
161 |
162 | /* get thread information*/
163 | thread_id = omp_get_thread_num();
164 | num_threads = omp_get_num_threads();
165 |
166 | /*initialise independent MT RNG stream for this thread*/
167 | vslNewStream(&stream_sims,VSL_BRNG_MT2203+thread_id,seed);
168 |
169 | /* distribute work among threads*/
170 | #pragma omp for schedule(guided)
171 | for (int i=0;i D(t)) where
209 | * D(t|\lambda) = 1/p(t|\lambda) and D(t) = 1/p_base(t)
210 | * p(t|\lambda) and p_base(t) are prior predictive distributions under the
211 | * priors p(\theta | \lambda) and base prior p(\theta) respectively.
212 | *
213 | * @param stream state pointer to this threads RNG
214 | * @param sigma0, sigma1 hyperparameters to test for weak informativity
215 | * @param num_datasets number of datasets to compute p-values over
216 | * @param d known parameters of forwards simulationn
217 | * @param p pointer to array to store p-value for each hyperparameter
218 | *
219 | * @note numerical overflow/underflow is avoided through computing with logs,
220 | * that is we actually compute Pr(log D(t|\lambda) < log D(t))
221 | */
222 | void
223 | compute_pvalues(VSLStreamStatePtr stream,double sigma0, double sigma1,
224 | unsigned int num_datasets, double * restrict d,
225 | double * restrict p)
226 | {
227 | __declspec(align(ALIGN)) int y[SIZE_D];
228 | double *theta;
229 | double *log_evidences_base;
230 | double *log_evidences_prior;
231 |
232 | /*allocate aligned memory for prior samples and evidences*/
233 | theta = (double*)_mm_malloc(2*num_datasets*sizeof(double),ALIGN);
234 | log_evidences_base = (double*)_mm_malloc(num_datasets*sizeof(double),ALIGN);
235 | log_evidences_prior = (double*)_mm_malloc(num_datasets*sizeof(double),ALIGN);
236 |
237 | /*compute evidences
238 | * under proposad prior based on data simulated from the base prior
239 | */
240 | vdRngGaussian(VSL_RNG_METHOD_GAUSSIAN_BOXMULLER2,stream,num_datasets,
241 | theta,0,SIGMA0_BASE);
242 | vdRngGaussian(VSL_RNG_METHOD_GAUSSIAN_BOXMULLER2,stream,num_datasets,
243 | theta+num_datasets,0,SIGMA1_BASE);
244 |
245 | /* generate data sets and compute evidences using SMC */
246 | for (unsigned int i=0;i= N/2*/
674 | /*compute normalised weight of large step in gamma*/
675 | #pragma omp simd
676 | for (unsigned int i=0;i w_max) ? w[i] : w_max;
686 | }
687 | w_sum = 0;
688 | #pragma omp simd reduction(+:w_sum)
689 | for (unsigned int i=0;i= ((double)N)/2.0)
711 | {
712 | gammavar_tp1 = 1.0;
713 | }
714 | else
715 | {
716 | /*smaller temperature step required*/
717 | double a = 0;
718 | double b = 0;
719 | double p = 0;
720 | double fa = 0;
721 | double fp = 0;
722 | double err = 0;
723 |
724 | /* find optimum step using bisection method*/
725 | a = gammavar_t + 1e-6;
726 | b = 1.0;
727 | p = (a+b)/2.0;
728 | fp = compute_ESS_diff(p,gammavar_t,w_temp,loglike,N);
729 | err = fabs(fp);
730 | while (err > 1e-5)
731 | {
732 | fa = compute_ESS_diff(a,gammavar_t,w_temp,loglike,N);
733 | if (fa*fp < 0)
734 | {
735 | b = p;
736 | }
737 | else
738 | {
739 | a = p;
740 | }
741 | p = (a+b)/2.0;
742 | fp = compute_ESS_diff(p,gammavar_t,w_temp,loglike,N);
743 | err = fabs(fp);
744 | }
745 | gammavar_tp1 = p;
746 | }
747 |
748 | /*re-weighting particles*/
749 | #pragma omp simd
750 | for (unsigned int i=0;i w_sum3 && k < N-1)
786 | {
787 | k++;
788 | w_sum3 += w[k];
789 | }
790 | r[i] = k;
791 | }
792 |
793 | /* re-assign particles*/
794 | for (unsigned int i=0;i u[i]);
881 | }
882 | expected_acc_probs = ((double)sum_acc_probs)/((double)N);
883 | if (expected_acc_probs <= 0.0)
884 | {
885 | expected_acc_probs =1e-6;
886 | }
887 | else if (expected_acc_probs >= 1.0)
888 | {
889 | expected_acc_probs = 1.0 - 1e-6;
890 | }
891 |
892 | /* determine the number of MCMC steps to ensure probability of 1-c that
893 | * a particle moves
894 | */
895 | R_t = (unsigned int)ceil(log(0.01)/log(1.0 - expected_acc_probs ));
896 |
897 | /*perform remaining repeats in blocks of VECLEN*/
898 | for (unsigned int i=0;i= 0 && x[j] > A)
1051 | {
1052 | x[j+1] = x[j];
1053 | j--;
1054 | }
1055 | x[j+1] = A;
1056 | }
1057 | }
1058 |
1059 |
--------------------------------------------------------------------------------